!14553 [GraphKernel] refine cast matmul fusion

From: @lingyunli63
Reviewed-by: @gaoxiong1,@ckey_dou
Signed-off-by: @ckey_dou

mindspore/ccsrc/backend/optimizer/graph_kernel/basic_ops_fusion.cc

@@ -39,20 +39,6 @@
 namespace mindspore {
 namespace opt {
 namespace {
-bool IsFusibleOp(const AnfNodePtr &node) {
-#if ENABLE_D
-  const std::set<std::string> graph_kernel_black_list = {"BNTrainingUpdateSum", "ApplyMomentum", "LayerNormForward",
-                                                         "LambNextMV", "LambUpdateWithLR"};
-  if (AnfAlgo::IsGraphKernel(node)) {
-    auto fg_attr = AnfAlgo::GetCNodeFuncGraphPtr(node)->get_attr(FUNC_GRAPH_ATTR_GRAPH_KERNEL);
-    if (fg_attr != nullptr) {
-      return graph_kernel_black_list.count(GetValue<std::string>(fg_attr)) == 0;
-    }
-  }
-#endif
-  return IsBasicFuseOp(node) || AnfAlgo::IsGraphKernel(node);
-}
-
 IncludeType IncludeFusedBasicOpForward(const AnfNodePtr &cur_node, const AnfNodePtr &node) {
   if (cur_node == node) {
     return FOLLOW;

mindspore/ccsrc/backend/optimizer/graph_kernel/cast_matmul_fusion.cc

@@ -0,0 +1,120 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "backend/optimizer/graph_kernel/cast_matmul_fusion.h"
+#include <tuple>
+#include "backend/session/anf_runtime_algorithm.h"
+#include "backend/kernel_compiler/common_utils.h"
+#include "backend/optimizer/graph_kernel/graph_kernel_helper.h"
+
+namespace mindspore {
+namespace opt {
+namespace {
+// Check if leaf is used by root
+bool HasPath(const AnfNodePtr &leaf, const AnfNodePtr &root, const FuncGraphManagerPtr &mng) {
+  MS_EXCEPTION_IF_NULL(mng);
+  bool result = false;
+  auto IncludeUser = [&result, &root](const AnfNodePtr &node) {
+    if (node == root) {
+      result = true;
+    }
+    return result ? EXCLUDE : FOLLOW;
+  };
+  static_cast<void>(DeepLinkedGraphSearch(leaf, IncludeUser));
+  return result;
+}
+}  // namespace
+
+/* MatMul supports fp32 bias, so remove the redundant cast if cast cannot fuse forword
+ *  case1, cast only used by MatMul
+ *
+ *   bias_fp32 = depend(bias_fp32, u)
+ *   %0 = cast(bias_fp32, fp16)
+ *   %1 = MatMul(A_fp16, B_fp16, %0)
+ *   ------>
+ *   bias_fp32 = depend(bias_fp32, u)
+ *   %1 = MatMul(A_fp16, B_fp16, bias_fp32)
+ *
+ *  case2, cast used by MatMul and UpdateStatus
+ *
+ *   bias_fp32 = load(p, status)
+ *   %0 = cast(bias_fp32, fp16)
+ *   %1 = MatMul(A_fp16, B_fp16, %0)
+ *   %2 = UpstateStatus(status, %0)
+ *   ------>
+ *   bias_fp32 = load(p, status)
+ *   %1 = MatMul(A_fp16, B_fp16, bias_fp32)
+ *   %2 = UpstateStatus(status, %1)
+ */
+bool CastMatmulFusion::Run(const FuncGraphPtr &func_graph) {
+  MS_EXCEPTION_IF_NULL(func_graph);
+  auto mng = func_graph->manager();
+  if (mng == nullptr) {
+    mng = Manage(func_graph, true);
+    func_graph->set_manager(mng);
+  }
+  auto changed = false;
+  auto nodes = TopoSort(func_graph->get_return());
+  for (auto node : nodes) {
+    if (!IsPrimitiveCNode(node, prim::kPrimMatMul)) {
+      continue;
+    }
+    auto cnode = node->cast<CNodePtr>();
+    if (cnode->size() != 4) {
+      continue;
+    }
+    auto cast_node = cnode->input(3);
+    if (!IsPrimitiveCNode(cast_node, prim::kPrimCast)) {
+      continue;
+    }
+    auto cast_input_type = AnfAlgo::GetInputDeviceDataType(cast_node, 0);
+    auto cast_output_type = AnfAlgo::GetOutputDeviceDataType(cast_node, 0);
+    if (cast_input_type != kNumberTypeFloat32 || cast_output_type != kNumberTypeFloat16) {
+      continue;
+    }
+    // Cast cannot fuse with its input
+    if (IsFusibleOp((cast_node->cast<CNodePtr>())->input(1))) {
+      continue;
+    }
+
+    auto user_index_set = mng->node_users()[cast_node];
+    // Case1 : Cast is only used by matmul
+    if (user_index_set.size() == 1) {
+      mng->Replace(cast_node, (cast_node->cast<CNodePtr>())->input(1));
+      changed = true;
+      continue;
+    }
+
+    // Case2 : Cast is used by matmul and Upstatus
+    if (user_index_set.size() > 2) {
+      continue;
+    }
+    for (auto user_index : user_index_set) {
+      // Exclude when UpdateStatus-> ... ->matmul path is found
+      if (IsPrimitiveCNode(user_index.first, prim::kPrimUpdateState) && !HasPath(user_index.first, node, mng)) {
+        auto update_state = (user_index.first)->cast<CNodePtr>();
+        update_state->set_input(2, node);
+        cnode->set_input(4, (cast_node->cast<CNodePtr>())->input(1));
+        mng->RemoveRoots();
+        mng->KeepRoots({func_graph});
+        changed = true;
+      }
+    }
+  }
+
+  return changed;
+}
+}  // namespace opt
+}  // namespace mindspore

mindspore/ccsrc/backend/optimizer/graph_kernel/optimize_matmul.h → mindspore/ccsrc/backend/optimizer/graph_kernel/cast_matmul_fusion.h

@@ -13,8 +13,8 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-#ifndef MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_OPTIMIZE_MATMUL_H_
-#define MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_OPTIMIZE_MATMUL_H_
+#ifndef MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_CAST_MATMUL_FUSION_H_
+#define MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_CAST_MATMUL_FUSION_H_
 
 #include <map>
 #include <memory>
@@ -24,13 +24,13 @@
 
 namespace mindspore {
 namespace opt {
-class OptimizeMatmul : public Pass {
+class CastMatmulFusion : public Pass {
  public:
-  OptimizeMatmul() : Pass("optimize_matmul") {}
-  ~OptimizeMatmul() override = default;
+  CastMatmulFusion() : Pass("cast_matmul_fusion") {}
+  ~CastMatmulFusion() override = default;
   bool Run(const FuncGraphPtr &graph) override;
 };
-using OptimizeMatmulPtr = std::shared_ptr<OptimizeMatmul>;
+using OptimizeMatmulPtr = std::shared_ptr<CastMatmulFusion>;
 }  // namespace opt
 }  // namespace mindspore
-#endif  // MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_OPTIMIZE_MATMUL_H_
+#endif  // MINDSPORE_CCSRC_BACKEND_OPTIMIZER_GRAPH_KERNEL_CAST_MATMUL_FUSION_H_

mindspore/ccsrc/backend/optimizer/graph_kernel/graph_kernel_helper.cc

@@ -627,6 +627,20 @@ bool IsBasicFuseOp(const AnfNodePtr &node) {
                      [&node](const PrimitivePtr &prim) { return IsPrimitiveCNode(node, prim); });
 }
 
+bool IsFusibleOp(const AnfNodePtr &node) {
+#if ENABLE_D
+  const std::set<std::string> graph_kernel_black_list = {"BNTrainingUpdateSum", "ApplyMomentum", "LayerNormForward",
+                                                         "LambNextMV", "LambUpdateWithLR"};
+  if (AnfAlgo::IsGraphKernel(node)) {
+    auto fg_attr = AnfAlgo::GetCNodeFuncGraphPtr(node)->get_attr(FUNC_GRAPH_ATTR_GRAPH_KERNEL);
+    if (fg_attr != nullptr) {
+      return graph_kernel_black_list.count(GetValue<std::string>(fg_attr)) == 0;
+    }
+  }
+#endif
+  return IsBasicFuseOp(node) || AnfAlgo::IsGraphKernel(node);
+}
+
 void ResetKernelInfo(const AnfNodePtr &node, KernelType kernel_type) {
   auto cnode = node->cast<CNodePtr>();
   MS_EXCEPTION_IF_NULL(cnode);

mindspore/ccsrc/backend/optimizer/graph_kernel/graph_kernel_helper.h

@@ -73,6 +73,7 @@ FuncGraphPtr JsonDescToAnf(const std::string &json_desc, const std::vector<AnfNo
 std::string ExtractGraphKernelName(const AnfNodePtrList &cnodes, const string &prefix = "", const string &postfix = "");
 std::vector<PrimitivePtr> GetFusibleOpList();
 bool IsBasicFuseOp(const AnfNodePtr &node);
+bool IsFusibleOp(const AnfNodePtr &node);
 void ResetKernelInfo(const AnfNodePtr &node, KernelType kernel_type = KernelType::UNKNOWN_KERNEL_TYPE);
 void InitDependPrior(const std::vector<AnfNodePtr> &todos,
                      std::multimap<AnfNodePtr, std::pair<AnfNodePtr, AnfNodePtr>> *depend_prior);

mindspore/ccsrc/backend/optimizer/graph_kernel/graph_kernel_optimization.cc

@@ -30,7 +30,7 @@
 #include "backend/optimizer/graph_kernel/tensor_promotion.h"
 #include "backend/optimizer/graph_kernel/graph_kernel_splitter.h"
 #include "backend/optimizer/graph_kernel/graph_kernel_expander.h"
-#include "backend/optimizer/graph_kernel/optimize_matmul.h"
+#include "backend/optimizer/graph_kernel/cast_matmul_fusion.h"
 #include "backend/optimizer/graph_kernel/raise_reduction_precision.h"
 #include "backend/optimizer/graph_kernel/graph_kernel_cse.h"
 #include "backend/optimizer/graph_kernel/shape_ops_splitter.h"
@@ -52,7 +52,7 @@ PassManagerPtr GraphKernelOptimizer::PreProcess() {
 
   if (is_ascend) {
     // Remove redundant Cast(bias, fp16) for Matmul input
-    pm->AddPass(std::make_shared<OptimizeMatmul>());
+    pm->AddPass(std::make_shared<CastMatmulFusion>());
 
     // Reorder TransData-Cast to Cast-TransData
     pm->AddPass(std::make_shared<ReorderOps>());

mindspore/ccsrc/backend/optimizer/graph_kernel/optimize_matmul.cc

@@ -1,64 +0,0 @@
-/**
- * Copyright 2021 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "backend/optimizer/graph_kernel/optimize_matmul.h"
-#include <tuple>
-#include "backend/session/anf_runtime_algorithm.h"
-#include "backend/kernel_compiler/common_utils.h"
-#include "backend/optimizer/graph_kernel/graph_kernel_helper.h"
-
-namespace mindspore {
-namespace opt {
-/* MatMul supports fp32 bias, so remove the redundant cast when cast only used by MatMul
- *
- *   %0 = cast(bias_fp32, fp16)
- *   %1 = MatMul(A_fp16, B_fp16, %0)
- *   ------>
- *   %1 = MatMul(A_fp16, B_fp16, bias_fp32)
- */
-bool OptimizeMatmul::Run(const FuncGraphPtr &func_graph) {
-  MS_EXCEPTION_IF_NULL(func_graph);
-  auto mng = func_graph->manager();
-  if (mng == nullptr) {
-    mng = Manage(func_graph, true);
-    func_graph->set_manager(mng);
-  }
-  auto changed = false;
-  auto nodes = TopoSort(func_graph->get_return());
-  for (auto node : nodes) {
-    if (!IsPrimitiveCNode(node, prim::kPrimMatMul)) {
-      continue;
-    }
-    auto cnode = node->cast<CNodePtr>();
-    if (cnode->size() != 4) {
-      continue;
-    }
-    auto cast_node = cnode->input(3);
-    if (!IsPrimitiveCNode(cast_node, prim::kPrimCast)) {
-      continue;
-    }
-    auto cast_input_type = AnfAlgo::GetInputDeviceDataType(cast_node, 0);
-    auto cast_output_type = AnfAlgo::GetOutputDeviceDataType(cast_node, 0);
-    if (cast_input_type == kNumberTypeFloat32 && cast_output_type == kNumberTypeFloat16 &&
-        mng->node_users()[cast_node].size() == 1) {
-      mng->Replace(cast_node, (cast_node->cast<CNodePtr>())->input(1));
-      changed = true;
-    }
-  }
-
-  return changed;
-}
-}  // namespace opt
-}  // namespace mindspore