Migrate mod op evaluate (openvinotoolkit#19687)

src/core/dev_api/shape_util.hpp

@@ -27,10 +27,55 @@ OPENVINO_API Shape make_dynamic_shape();
 OPENVINO_DEPRECATED("This function is deprecated and will be removed soon.")
 OPENVINO_API bool is_dynamic_shape(const Shape& s);
 
+/**
+ * @brief Creates reduced shape from input by removing dimensions.
+ *
+ * @param input      Input shape for reduce calculation.
+ * @param axes   Reduction axes.
+ * @return Reduced shape.
+ */
 OPENVINO_API Shape reduce(const Shape& input, const AxisSet& axes);
+
+/**
+ * @brief Creates reduced shape from input removing or replacing dimension.
+ *
+ * The reduction type depends on `keep_dims` flags. If it's set to true then reduced dimension will be replaced by `1`,
+ * otherwise removed.
+ *
+ * @param input      Input shape for reduce calculation.
+ * @param axes       Reduction axes.
+ * @param keep_dims  Flag to keep reduced dimension.
+ * @return Reduced shape.
+ */
 OPENVINO_API Shape reduce(const Shape& input, const AxisSet& axes, const bool keep_dims);
+
+/**
+  * @brief Creates reduced vector from input by removing elements.
+ *
+ * @param input  Input vector for reduce calculation.
+ * @param axes   Reduction axes.
+ * @return Reduced vector
+ */
 OPENVINO_API std::vector<size_t> reduce(const std::vector<size_t>& input, const AxisSet& axes);
 
+/**
+ * @brief Creates reduced shape from input by replacing reduced dimension with `1`.
+ *
+ * @param input      Input shape for reduce calculation.
+ * @param axes       Reduction axes.
+ * @return Reduced shape.
+ */
 OPENVINO_API Shape reduce_keep_dims(const Shape& input, const AxisSet& axes);
+
+/**
+ * @brief Get the broadcast shape as merge second shape into first according to broadcast specification.
+ *
+ * @param first           First input shape.
+ * @param second          Second input shape.
+ * @param broadcast_spec  Broadcast specification.
+ *
+ * @return Result shape from inputs with applied broadcast specification.
+ */
+Shape get_broadcast_shape(const Shape& first, const Shape& second, const op::AutoBroadcastSpec& broadcast_spec);
 }  // namespace util
 }  // namespace ov

src/core/reference/include/openvino/reference/mod.hpp

@@ -8,19 +8,41 @@
 #include <cstddef>
 
 #include "openvino/reference/autobroadcast_binop.hpp"
+#include "openvino/reference/utils/type_util.hpp"
 
 namespace ov {
 namespace reference {
-template <typename T>
-void mod(const T* arg0,
-         const T* arg1,
-         T* out,
+namespace func {
+template <class T, typename std::enable_if<std::is_integral<T>::value>::type* = nullptr>
+constexpr T mod(const T x, const T y) {
+    return x % y;
+}
+
+template <class T, typename std::enable_if<ov::is_floating_point<T>()>::type* = nullptr>
+T mod(const T x, const T y) {
+    return x - (std::trunc(x / y) * y);
+}
+}  // namespace func
+
+/**
+ * @brief Reference implementation of binary elementwise Mod operator.
+ *
+ * @param arg0            Iterator to input 0 data.
+ * @param arg1            Iterator to input 1 data.
+ * @param out             Iterator to output data.
+ * @param arg_shape0      Input 0 shape.
+ * @param arg_shape1      Input 1 shape.
+ * @param broadcast_spec  Broadcast specification mode.
+ */
+template <class InputIt, class OutputIt>
+void mod(InputIt arg0,
+         InputIt arg1,
+         OutputIt out,
          const Shape& arg_shape0,
          const Shape& arg_shape1,
          const op::AutoBroadcastSpec& broadcast_spec) {
-    autobroadcast_binop(arg0, arg1, out, arg_shape0, arg_shape1, broadcast_spec, [](T x, T y) -> T {
-        return static_cast<T>(x - std::truncf(static_cast<float>(x / y)) * y);
-    });
+    using T = typename std::iterator_traits<OutputIt>::value_type;
+    autobroadcast_binop(arg0, arg1, out, arg_shape0, arg_shape1, broadcast_spec, &func::mod<T>);
 }
 }  // namespace reference
 }  // namespace ov

src/core/src/op/mod.cpp

@@ -2,101 +2,79 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "ngraph/op/mod.hpp"
+#include "openvino/op/mod.hpp"
 
+#include "element_visitor.hpp"
 #include "itt.hpp"
 #include "openvino/reference/mod.hpp"
+#include "shape_util.hpp"
 
-using namespace std;
-using namespace ngraph;
+namespace ov {
+namespace op {
+namespace mod {
+struct Evaluate : ov::element::NoAction<bool> {
+    using ov::element::NoAction<bool>::visit;
 
-// ------------------------------ v1 -------------------------------------------
+    template <element::Type_t ET>
+    static result_type visit(const Tensor& in0,
+                             const Tensor& in1,
+                             Tensor& out,
+                             const AutoBroadcastSpec& broadcast_spec) {
+        using T = typename element_type_traits<ET>::value_type;
+        reference::mod(in0.data<const T>(),
+                       in1.data<const T>(),
+                       out.data<T>(),
+                       in0.get_shape(),
+                       in1.get_shape(),
+                       broadcast_spec);
+        return true;
+    }
+};
+}  // namespace mod
 
-op::v1::Mod::Mod(const Output<Node>& arg0, const Output<Node>& arg1, const AutoBroadcastSpec& auto_broadcast)
+namespace v1 {
+v1::Mod::Mod(const Output<Node>& arg0, const Output<Node>& arg1, const AutoBroadcastSpec& auto_broadcast)
     : BinaryElementwiseArithmetic(arg0, arg1, auto_broadcast) {
     constructor_validate_and_infer_types();
 }
 
-shared_ptr<Node> op::v1::Mod::clone_with_new_inputs(const OutputVector& new_args) const {
+std::shared_ptr<Node> Mod::clone_with_new_inputs(const OutputVector& new_args) const {
     OV_OP_SCOPE(v1_Mod_clone_with_new_inputs);
     check_new_args_count(this, new_args);
-    return make_shared<Mod>(new_args.at(0), new_args.at(1), this->get_autob());
-}
-
-namespace mod_op {
-namespace {
-template <typename T>
-bool evaluate(const ov::Tensor& arg0,
-              const ov::Tensor& arg1,
-              const ov::Tensor& out,
-              const op::AutoBroadcastSpec& broadcast_spec) {
-    ov::reference::mod(arg0.data<T>(),
-                       arg1.data<T>(),
-                       out.data<T>(),
-                       arg0.get_shape(),
-                       arg1.get_shape(),
-                       broadcast_spec);
-    return true;
+    return std::make_shared<Mod>(new_args.at(0), new_args.at(1), this->get_autob());
 }
 
-bool evaluate_mod(const ov::Tensor& arg0,
-                  const ov::Tensor& arg1,
-                  const ov::Tensor& out,
-                  const op::AutoBroadcastSpec& broadcast_spec) {
-    bool rc = true;
-    switch (arg0.get_element_type()) {
-    case ov::element::Type_t::i8: {
-        rc = evaluate<int8_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::i16: {
-        rc = evaluate<int16_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::i32: {
-        rc = evaluate<int32_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::i64: {
-        rc = evaluate<int64_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::u8: {
-        rc = evaluate<uint8_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::u16: {
-        rc = evaluate<uint16_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::u32: {
-        rc = evaluate<uint32_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    case ov::element::Type_t::u64: {
-        rc = evaluate<uint64_t>(arg0, arg1, out, broadcast_spec);
-    } break;
-    default:
-        rc = false;
-        break;
-    }
-    return rc;
-}
-}  // namespace
-}  // namespace mod_op
-
-bool op::v1::Mod::evaluate(ov::TensorVector& outputs, const ov::TensorVector& inputs) const {
+bool Mod::evaluate(ov::TensorVector& outputs, const ov::TensorVector& inputs) const {
     OV_OP_SCOPE(v1_Mod_evaluate);
-    return mod_op::evaluate_mod(inputs[0], inputs[1], outputs[0], get_autob());
+    OPENVINO_ASSERT(outputs.size() == 1);
+    OPENVINO_ASSERT(inputs.size() == 2);
+
+    outputs[0].set_shape(ov::util::get_broadcast_shape(inputs[0].get_shape(), inputs[1].get_shape(), get_autob()));
+    using namespace ov::element;
+    return IfTypeOf<i8, i16, i32, i64, u8, u16, u32, u64>::apply<mod::Evaluate>(inputs[0].get_element_type(),
+                                                                                inputs[0],
+                                                                                inputs[1],
+                                                                                outputs[0],
+                                                                                get_autob());
 }
 
-bool op::v1::Mod::has_evaluate() const {
+bool Mod::has_evaluate() const {
     OV_OP_SCOPE(v1_Mod_has_evaluate);
+
     switch (get_input_element_type(0)) {
-    case ngraph::element::i8:
-    case ngraph::element::i16:
-    case ngraph::element::i32:
-    case ngraph::element::i64:
-    case ngraph::element::u8:
-    case ngraph::element::u16:
-    case ngraph::element::u32:
-    case ngraph::element::u64:
+    case element::i8:
+    case element::i16:
+    case element::i32:
+    case element::i64:
+    case element::u8:
+    case element::u16:
+    case element::u32:
+    case element::u64:
         return true;
     default:
-        break;
+        return false;
     }
-    return false;
 }
+}  // namespace v1
+}  // namespace op
+}  // namespace ov

src/core/src/shape_util.cpp

@@ -6,6 +6,7 @@
 
 #include <algorithm>
 
+#include "openvino/core/partial_shape.hpp"
 #include "shape_util.hpp"
 
 using namespace ngraph;
@@ -128,5 +129,12 @@ std::vector<size_t> reduce(const std::vector<size_t>& input, const AxisSet& axes
 Shape reduce_keep_dims(const Shape& input, const AxisSet& axes) {
     return ov::replace_container(input, axes);
 }
+
+Shape get_broadcast_shape(const Shape& first, const Shape& second, const op::AutoBroadcastSpec& broadcast_spec) {
+    auto out_shape = PartialShape(first);
+    OPENVINO_ASSERT(PartialShape::broadcast_merge_into(out_shape, second, broadcast_spec),
+                    "Argument shapes are inconsistent");
+    return out_shape.to_shape();
+}
 }  // namespace util
 }  // namespace ov

src/plugins/template/backend/ops/mod.cpp

@@ -11,12 +11,12 @@ bool evaluate(const std::shared_ptr<ngraph::op::v1::Mod>& op,
               const ngraph::HostTensorVector& outputs,
               const ngraph::HostTensorVector& inputs) {
     using T = typename ngraph::element_type_traits<ET>::value_type;
-    ov::reference::mod<T>(inputs[0]->get_data_ptr<T>(),
-                          inputs[1]->get_data_ptr<T>(),
-                          outputs[0]->get_data_ptr<T>(),
-                          inputs[0]->get_shape(),
-                          inputs[1]->get_shape(),
-                          op->get_autob());
+    ov::reference::mod(inputs[0]->get_data_ptr<const T>(),
+                       inputs[1]->get_data_ptr<const T>(),
+                       outputs[0]->get_data_ptr<T>(),
+                       inputs[0]->get_shape(),
+                       inputs[1]->get_shape(),
+                       op->get_autob());
     return true;
 }