WIP: Generic Reorder

antonvor · Oct 12, 2023 · 5dcc4f7 · 5dcc4f7
1 parent 749ed9d
commit 5dcc4f7
Show file tree

Hide file tree

Showing 12 changed files with 675 additions and 19 deletions.
diff --git a/src/plugins/intel_cpu/src/dnnl_extension_utils.cpp b/src/plugins/intel_cpu/src/dnnl_extension_utils.cpp
@@ -36,6 +36,9 @@ uint8_t DnnlExtensionUtils::sizeOfDataType(dnnl::memory::data_type dataType) {
         return 2;
     case dnnl::memory::data_type::undef:
         return 0;
+    // todo:
+    case dnnl::memory::data_type::f64:
+        return 8;
     default:
         IE_THROW() << "Unsupported data type.";
     }
@@ -60,6 +63,8 @@ memory::data_type DnnlExtensionUtils::IEPrecisionToDataType(const InferenceEngin
             return memory::data_type::f16;
         case InferenceEngine::Precision::UNSPECIFIED:
             return memory::data_type::undef;
+        case InferenceEngine::Precision::FP64:
+            return memory::data_type::f64;
         default: {
             IE_THROW() << "The plugin does not support " << prec.name();
         }
@@ -84,6 +89,9 @@ InferenceEngine::Precision DnnlExtensionUtils::DataTypeToIEPrecision(memory::dat
             return InferenceEngine::Precision::FP16;
         case memory::data_type::undef:
             return InferenceEngine::Precision::UNSPECIFIED;
+        // todo: needed for graph dumping, remove later
+        case memory::data_type::f64:
+            return InferenceEngine::Precision::FP64;
         default: {
             IE_THROW() << "Unsupported data type.";
         }

diff --git a/src/plugins/intel_cpu/src/graph.cpp b/src/plugins/intel_cpu/src/graph.cpp
@@ -415,9 +415,9 @@ void Graph::InitDescriptors() {
         DEBUG_LOG("Init supported primitive descriptors for node: ", node->getName());
         node->initSupportedPrimitiveDescriptors();
 
-        OV_ITT_SCOPE_NEXT(FIRST_INFERENCE, taskChain, node->profiling.filterSupportedPrimitiveDescriptors);
-        DEBUG_LOG("Filter supported primitive descriptors for node: ", node->getName());
-        node->filterSupportedPrimitiveDescriptors();
+        // OV_ITT_SCOPE_NEXT(FIRST_INFERENCE, taskChain, node->profiling.filterSupportedPrimitiveDescriptors);
+        // DEBUG_LOG("Filter supported primitive descriptors for node: ", node->getName());
+        // node->filterSupportedPrimitiveDescriptors();
 
 #ifdef CPU_DEBUG_CAPS
         const auto& SPDs = node->getSupportedPrimitiveDescriptors();
@@ -599,6 +599,11 @@ void Graph::InitEdges() {
         auto reorderStatus = graphEdges[i]->needReorder();
         DEBUG_LOG(graphEdges[i]->name(), " reorderStatus = ", reorderStatus);
         if (reorderStatus == Edge::ReorderStatus::Regular) {
+            if (edge->getParent()->getName() == "TRANSPOSE_1") {
+                insertReorder(edge, false);
+                updateEdge(i);
+                continue;
+            }
             Edge::ReorderStatus reorderStatusInternal = Edge::ReorderStatus::Regular;
             // Check if there is a reorder that needs the precision conversion
             if (edge->getInputDesc().getPrecision() != edge->getOutputDesc().getPrecision() &&

diff --git a/src/plugins/intel_cpu/src/graph_optimizer.cpp b/src/plugins/intel_cpu/src/graph_optimizer.cpp
@@ -99,6 +99,10 @@ void GraphOptimizer::ApplyCommonGraphOptimizations(Graph &graph) {
     FuseFCAndTransposeOnWeights(graph);
     graph.RemoveDroppedNodes();
 
+    OV_ITT_SCOPE_NEXT(FIRST_INFERENCE, taskChain, "RemoveConvert");
+    RemoveConvert(graph);
+    graph.RemoveDroppedNodes();
+
     OV_ITT_SCOPE_NEXT(FIRST_INFERENCE, taskChain, "FuseDeconvolutionAndSimpleOperation");
     FuseDeconvolutionAndSimpleOperation(graph);
     graph.RemoveDroppedNodes();
@@ -193,8 +197,8 @@ void GraphOptimizer::ApplyImplSpecificGraphOptimizations(Graph &graph) {
     DropDoubleReorders(graph);
     graph.RemoveDroppedNodes();
 
-    MergeTransposeAndReorder(graph);
-    graph.RemoveDroppedNodes();
+    // MergeTransposeAndReorder(graph);
+    // graph.RemoveDroppedNodes();
 
     graph.RemoveDroppedEdges();
 }
@@ -848,6 +852,23 @@ void GraphOptimizer::FuseFCAndConvertOnWeights(Graph& graph) {
     }
 }
 
+void GraphOptimizer::RemoveConvert(Graph& graph) {
+    auto& graphNodes = graph.GetNodes();
+
+    auto isSuitablePattern = [](NodePtr parent) {
+        bool res = true && parent->getType() == Type::Convert
+            && parent->getParentEdgeAt(0)->getParent()->getName() == "TRANSPOSE_1";
+        return res;
+    };
+
+    for (auto parent : graphNodes) {
+        if (isSuitablePattern(parent)) {
+            CPU_GRAPH_OPTIMIZER_SCOPE(RemoveConvert);
+            graph.DropNode(parent);
+        }
+    }
+}
+
 void GraphOptimizer::FuseFCAndTransposeOnWeights(Graph& graph) {
     // This optimization allows us to avoid transposing the weights in Transpose node and do it directly along with reordering in FC node
     auto& graphNodes = graph.GetNodes();

diff --git a/src/plugins/intel_cpu/src/graph_optimizer.h b/src/plugins/intel_cpu/src/graph_optimizer.h
@@ -27,6 +27,7 @@ class GraphOptimizer {
     void FuseMultiplyAndAdd(Graph &graph);
     void MergeConvertAndScaleShift(Graph& graph);
     void FuseFCAndConvertOnWeights(Graph& graph);
+    void RemoveConvert(Graph &graph);
     void FuseFCAndTransposeOnWeights(Graph& graph);
     void FuseFullyConnectedAndSimpleOperation(Graph &graph);
     void FuseMatMulAndSimpleOperation(Graph &graph);