close BatchMatmul and ReduceSum in graph kernel

mindspore/ccsrc/backend/kernel_compiler/kernel_query.cc

@@ -105,10 +105,6 @@ void KernelQuery(const CNodePtr &kernel_node, std::vector<std::shared_ptr<kernel
 
   auto context_ptr = MsContext::GetInstance();
   MS_EXCEPTION_IF_NULL(context_ptr);
-  if (context_ptr->get_param<bool>(MS_CTX_ENABLE_GRAPH_KERNEL) &&
-      IsPrimitiveCNode(kernel_node, prim::kPrimBatchMatMul)) {
-    kernel_type = KernelType::AKG_KERNEL;
-  }
 
   switch (kernel_type) {
     case KernelType::AKG_KERNEL:

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

@@ -736,11 +736,11 @@ std::string ExtractGraphKernelName(const AnfNodePtrList &cnodes, const string &p
 std::vector<PrimitivePtr> GetFusibleOpList() {
 #if ENABLE_D
   std::vector<PrimitivePtr> fusible_basic_ops = {
-    prim::kPrimAbs,        prim::kPrimRound,      prim::kPrimNeg,       prim::kPrimExp,     prim::kPrimTensorAdd,
-    prim::kPrimExpandDims, prim::kPrimMul,        prim::kPrimMinimum,   prim::kPrimMaximum, prim::kPrimLog,
-    prim::kPrimPow,        prim::kPrimSub,        prim::kPrimRsqrt,     prim::kPrimSqrt,    prim::kPrimAddN,
-    prim::kPrimEqual,      prim::kPrimReciprocal, prim::kPrimReduceSum, prim::kPrimTanh,    prim::kPrimReshape,
-    prim::kPrimTranspose,  prim::kPrimCast,       prim::kPrimRealDiv};
+    prim::kPrimAbs,        prim::kPrimRound,      prim::kPrimNeg,     prim::kPrimExp,     prim::kPrimTensorAdd,
+    prim::kPrimExpandDims, prim::kPrimMul,        prim::kPrimMinimum, prim::kPrimMaximum, prim::kPrimLog,
+    prim::kPrimPow,        prim::kPrimSub,        prim::kPrimRsqrt,   prim::kPrimSqrt,    prim::kPrimAddN,
+    prim::kPrimEqual,      prim::kPrimReciprocal, prim::kPrimTanh,    prim::kPrimReshape, prim::kPrimTranspose,
+    prim::kPrimCast,       prim::kPrimRealDiv};
 #elif ENABLE_GPU
   std::vector<PrimitivePtr> fusible_basic_ops = {
     prim::kPrimAbs,     prim::kPrimRound,      prim::kPrimNeg,       prim::kPrimExp,     prim::kPrimTensorAdd,