!7556 Optimize the performance of function ProcessVmArg

Merge pull request !7556 from chenfei_mindspore/ProcessVmArgPerformanceOptimize

mindspore/ccsrc/pipeline/jit/base.h

@@ -31,7 +31,10 @@ namespace pipeline {
 struct ExecutorInfo {
   FuncGraphPtr func_graph;
   ResourcePtr resource;
+  // The num of input data.
   std::size_t arg_list_size;
+  // The all args of graph,including input data and weight.
+  VectorRef arg_list;
 };
 using ExecutorInfoPtr = std::shared_ptr<ExecutorInfo>;
 

mindspore/ccsrc/pipeline/jit/pipeline.cc

@@ -181,13 +181,6 @@ FuncGraphPtr ExecutorPy::GetFuncGraph(const std::string &phase) {
   return info_[phase]->func_graph;
 }
 
-std::size_t ExecutorPy::ArgListSize(const std::string &phase) {
-  if (info_.count(phase) == 0) {
-    MS_LOG(EXCEPTION) << "No phase in executor:" << GetPhasePrefix(phase);
-  }
-  return info_[phase]->arg_list_size;
-}
-
 compile::VmEvalFuncPtr ExecutorPy::GetVmEvalFunc(const std::string &phase) {
   ResourcePtr res = GetResource(phase);
   MS_EXCEPTION_IF_NULL(res);
@@ -702,8 +695,9 @@ void Pipeline::Run() {
 }
 
 void ProcessVmArgInner(const py::tuple &args, const ResourcePtr &res, VectorRef *const arg_list) {
+  MS_EXCEPTION_IF_NULL(arg_list);
   std::size_t size = args.size();
-
+  bool arg_list_inited = !arg_list->empty();
   for (std::size_t i = 0; i < size; i++) {
     py::object arg = args[i];
     auto ms_context = MsContext::GetInstance();
@@ -715,7 +709,14 @@ void ProcessVmArgInner(const py::tuple &args, const ResourcePtr &res, VectorRef
     if (!succ) {
       MS_LOG(EXCEPTION) << "The " << i << "th arg convert failed.";
     }
-    arg_list->push_back(converted);
+    if (!arg_list_inited) {
+      arg_list->push_back(converted);
+      continue;
+    }
+    if (i >= arg_list->size()) {
+      MS_LOG(EXCEPTION) << "i:" << i << " output of range:" << arg_list->size();
+    }
+    (*arg_list)[i] = converted;
   }
 
   MS_EXCEPTION_IF_NULL(res);
@@ -792,20 +793,23 @@ py::object ExecutorPy::Run(const py::tuple &args, const py::object &phase) {
     return args;
   }
 #endif
-  std::size_t full_arg_size = ArgListSize(phase_s);
-  if (size > full_arg_size) {
-    MS_LOG(WARNING) << "The arg num : size = " << size << ". full_arg_size = " << full_arg_size;
+  auto iter = info_.find(phase_s);
+  if (iter == info_.end()) {
+    MS_LOG(EXCEPTION) << "No phase in executor:" << GetPhasePrefix(phase_s);
   }
-  VectorRef arg_list;
-  ProcessVmArg(args, phase_s, &arg_list);
-
+  auto &execute_info = iter->second;
+  MS_EXCEPTION_IF_NULL(execute_info);
+  if (size > execute_info->arg_list_size) {
+    MS_LOG(WARNING) << "The arg num : size = " << size << ". full_arg_size = " << execute_info->arg_list_size;
+  }
+  ProcessVmArg(args, phase_s, &execute_info->arg_list);
+  // Start to run phase.
   compile::VmEvalFuncPtr run = GetVmEvalFunc(phase_s);
   if (run == nullptr) {
     MS_LOG(EXCEPTION) << "Can't find run graph func for " << phase_s;
   }
-
   MS_LOG(DEBUG) << "Eval run" << backend;
-  BaseRef value = (*run)(arg_list);
+  BaseRef value = (*run)(execute_info->arg_list);
   MS_LOG(DEBUG) << "Run end";
   return BaseRefToPyData(value);
 }

mindspore/ccsrc/pipeline/jit/pipeline.h

@@ -82,7 +82,6 @@ class ExecutorPy : public std::enable_shared_from_this<ExecutorPy> {
   ResourcePtr GetResource(const std::string &phase);
   FuncGraphPtr GetFuncGraph(const std::string &phase);
   py::bytes GetFuncGraphProto(const std::string &phase, const std::string &type);
-  std::size_t ArgListSize(const std::string &phase);
   compile::VmEvalFuncPtr GetVmEvalFunc(const std::string &phase);
   bool HasCompiled(const std::string &phase) const;