[DEBUG]: BF16 validation

openvinotoolkit · Apr 9, 2023 · c1b16bf · c1b16bf
1 parent 6aff821
commit c1b16bf
Show file tree

Hide file tree

Showing 5 changed files with 33 additions and 2 deletions.
diff --git a/src/inference/src/cpp_interfaces/interface/ie_iplugin_internal.cpp b/src/inference/src/cpp_interfaces/interface/ie_iplugin_internal.cpp
@@ -136,8 +136,15 @@ std::map<std::string, std::shared_ptr<const T>> const_map_cast(const std::map<st
 
 std::shared_ptr<IExecutableNetworkInternal> IInferencePlugin::LoadNetwork(
     const CNNNetwork& orig_network,
-    const std::map<std::string, std::string>& config,
+    const std::map<std::string, std::string>& config_,
     const std::shared_ptr<RemoteContext>& context) {
+    auto config = config_;
+
+    // TODO: just to test
+    ov::element::Type hint = ov::element::bf16;
+    config.emplace(std::string("INFERENCE_PRECISION_HINT"), hint.get_type_name());
+    std::cout << "IInferencePlugin::LoadNetwork: INFERENCE_PRECISION_HINT: " << hint.get_type_name() << std::endl;
+
     std::shared_ptr<IExecutableNetworkInternal> impl;
 
     // if IR `version` is not set, suppose it's IR v10 for old API

diff --git a/src/plugins/intel_cpu/src/config.cpp b/src/plugins/intel_cpu/src/config.cpp
@@ -67,9 +67,17 @@ void Config::applyDebugCapsProperties() {
 }
 #endif
 
-void Config::readProperties(const std::map<std::string, std::string> &prop) {
+void Config::readProperties(const std::map<std::string, std::string> &prop_) {
     const auto streamExecutorConfigKeys = streamExecutorConfig.SupportedKeys();
     const auto hintsConfigKeys = perfHintsConfig.SupportedKeys();
+
+    auto prop = prop_;
+
+    // TODO: just to test
+    ov::element::Type hint = ov::element::bf16;
+    prop.emplace(std::string("INFERENCE_PRECISION_HINT"), hint.get_type_name());
+    std::cout << "Config::readProperties: INFERENCE_PRECISION_HINT: " << hint.get_type_name() << std::endl;
+
     for (const auto& kvp : prop) {
         const auto& key = kvp.first;
         const auto& val = kvp.second;

diff --git a/src/plugins/intel_cpu/src/graph.cpp b/src/plugins/intel_cpu/src/graph.cpp
@@ -192,6 +192,13 @@ void Graph::Replicate(const std::shared_ptr<const ov::Model> &subgraph) {
         graphNodes.push_back(outNode);
     }
 
+    std::cout << "Graph::Replicate" << std::endl;
+    if (dnnl::impl::cpu::x64::mayiuse(dnnl::impl::cpu::x64::avx512_core_bf16)) {
+        std::cout << "Graph::Replicate: BF16 is supported" << std::endl;
+    } else {
+        std::cout << "Graph::Replicate: BF16 is not supported" << std::endl;
+    }
+
     if (getConfig().enforceBF16)
         EnforceBF16();
 }
@@ -308,6 +315,12 @@ void Graph::Replicate(const CNNNetwork &network) {
         graphNodes.push_back(outNode);
     }
 
+    if (dnnl::impl::cpu::x64::mayiuse(dnnl::impl::cpu::x64::avx512_core_bf16)) {
+        std::cout << "Graph::Replicate: BF16 is supported" << std::endl;
+    } else {
+        std::cout << "Graph::Replicate: BF16 is not supported" << std::endl;
+    }
+
     if (getConfig().enforceBF16)
         EnforceBF16();
 

diff --git a/src/plugins/intel_cpu/src/nodes/subgraph.cpp b/src/plugins/intel_cpu/src/nodes/subgraph.cpp
@@ -548,6 +548,7 @@ void Snippet::generate(const jit_snippets_compile_args* jcp) {
         // Note, MatMul decomposition will be ran later again for case if BF16 enforcement is not happened
         pre_dialect.register_pass<ngraph::snippets::pass::MatMulToBrgemm>();
         pre_dialect.register_pass<pass::EnforcePrecision>(element::f32, element::bf16);
+        std::cout << "Snippet::generate: EnforcePrecision was added in pipeline" << std::endl;
     }
 
     ov::pass::Manager post_dialect;

diff --git a/src/plugins/intel_cpu/src/snippets_transformations/enforce_precision.cpp b/src/plugins/intel_cpu/src/snippets_transformations/enforce_precision.cpp
@@ -102,9 +102,11 @@ bool EnforcePrecision::run_on_model(const std::shared_ptr<ov::Model>& f) {
                     // remove convert
                     op_parent->output(0).replace(op_parent->get_input_source_output(0));
                     was_updated = true;
+                    std::cout << "EnforcePrecision::run_on_model: convertion removal: " << op->get_type_name() << ":" << op->get_friendly_name() << std::endl;
                 } else if (supported_precisions_to_enforce[index] != actual_precisions[index]) {
                     insert_convert(op->get_input_source_output(index), op, index, target);
                     was_updated = true;
+                    std::cout << "EnforcePrecision::run_on_model: convertion insertion: " << op->get_type_name() << ":" << op->get_friendly_name() << std::endl;
                 }
             }
         }