!1840 Add op pre compiler process

Merge pull request !1840 from wangcong/master

mindspore/_extends/parallel_compile/tbe_compiler/compiler.py

@@ -28,7 +28,8 @@ build_in_impl_path = get_build_in_impl_path()
 # op function list
 op_build = "compile"
 op_pre_build = "pre_build"
-
+fusion_type_map = {'Convolution': 0, 'ElemWise': 1, 'CommReduce': 2,
+                   'Segment': 3, 'Opaque': 4}
 
 def _initialize(impl_path):
     """Initialize"""
@@ -108,7 +109,7 @@ def build_op(build_type, json_str):
 
         # pre build
         if build_type == op_pre_build:
-            op_func(*inputs_args, *outputs_args, *attrs_args, kernel_name)
+            op_func(*inputs_args, *outputs_args, *attrs_args, kernel_name=kernel_name)
             # disable only pattern configuration
             op_build_cfg_en()
             return get_op_pattern()
@@ -159,11 +160,16 @@ def compile_with_json(json_str):
     json_info = json.loads(json_str)
     if "fusion_op" in json_info:
         ret = compile_fusion_op(json_str)
+    elif "compile_type" in json_info:
+        ret = build_op(op_pre_build, json_str)
     else:
         ret = build_op(op_build, json_str)
     return ret
 
-
 if __name__ == "__main__":
     in_args = sys.stdin.readline()
-    compile_with_json(in_args)
+    result = compile_with_json(in_args)
+    if result in fusion_type_map:
+        exit(fusion_type_map[result])
+    else:
+        exit(100)

mindspore/_extends/parallel_compile/tbe_compiler/tbe_process.py

@@ -75,7 +75,6 @@ def check_supported(op_json: str):
 
     return ret
 
-
 def run_compiler(op_json):
     """
     run compiler to compile op with subprocess
@@ -88,15 +87,16 @@ def run_compiler(op_json):
     """
     try:
         tbe_compiler = os.path.join(os.path.split(os.path.realpath(__file__))[0], "compiler.py")
-        subprocess.run([sys.executable, tbe_compiler], input=op_json, timeout=300,
-                       text=True, capture_output=True, check=True)
-        return "Success", "Success"
+        completed_object = subprocess.run([sys.executable, tbe_compiler], input=op_json, timeout=300,
+                                          text=True, capture_output=True, check=False)
+        if completed_object:
+            code = completed_object.returncode
+        return "Success", str(code)
     except subprocess.TimeoutExpired:
         tb = traceback.format_exc()
-        return "TBEException", "CompileTimeOut: " + tb + "\ninput_args: " + op_json
+        return "TBEException", "PreCompileTimeOut: " + tb + "\ninput_args: " + op_json
     except subprocess.CalledProcessError as e:
-        return "TBEException", "CompileProcessFailed:\n" + e.stdout + "\n" + e.stderr + "\ninput_args: " + op_json
-
+        return "TBEException", "PreCompileProcessFailed:\n" + e.stdout + "\n" + e.stderr + "\ninput_args: " + op_json
 
 class CompilerPool:
     """compiler pool"""
@@ -154,11 +154,11 @@ class CompilerPool:
             task_id, task_future = self.__running_tasks.pop(0)
             ret_type, result = task_future.get(330)
             if ret_type == "Success":
-                ret = task_id, "Success"
+                ret = task_id, "Success", result
             elif ret_type in ("Exception", "TBEException"):
-                ret = task_id, ret_type + ":" + result
+                ret = task_id, ret_type + ":" + result, "_"
             else:
-                ret = task_id, "Exception: Not support return type:" + str(ret_type)
+                ret = task_id, "Exception: Not support return type:" + str(ret_type), "_"
         return ret
 
     def reset_task_info(self):

mindspore/ccsrc/device/ascend/kernel_build_ascend.cc

@@ -62,6 +62,31 @@ static kernel::KernelModPtr SerialCompileImpl(const AnfNodePtr &anf_node) {
   return kernel_mod_ptr;
 }
 
+static bool KernelPreBuildParallelCompile(const mindspore::session::KernelGraph *kernel_graph_ptr) {
+  MS_EXCEPTION_IF_NULL(kernel_graph_ptr);
+  std::vector<AnfNodePtr> tbe_nodes;
+  for (const auto &anf_node : kernel_graph_ptr->execution_order()) {
+    MS_EXCEPTION_IF_NULL(anf_node);
+    if (!AnfAlgo::IsRealKernel(anf_node)) {
+      continue;
+    }
+    KernelType kernel_type = AnfAlgo::GetKernelType(anf_node);
+    switch (kernel_type) {
+      case KernelType::TBE_KERNEL: {
+        if (AnfAlgo::GetKernelMod(anf_node) == nullptr) {
+          tbe_nodes.push_back(anf_node);
+        }
+        break;
+      }
+      default: {
+        break;
+      }
+    }
+  }
+  bool ret = kernel::TbeOpParallelPreBuild(tbe_nodes);
+  return ret;
+}
+
 static bool KernelBuildParallelCompile(const mindspore::session::KernelGraph *kernel_graph_ptr) {
   MS_EXCEPTION_IF_NULL(kernel_graph_ptr);
   std::vector<AnfNodePtr> tbe_nodes;
@@ -188,6 +213,12 @@ bool IsAtomicNode(const CNodePtr &kernel_node) {
   return atomic_flag;
 }
 
+bool KernelPreBuild(const mindspore::session::KernelGraph *kernel_graph_ptr) {
+  MS_EXCEPTION_IF_NULL(kernel_graph_ptr);
+  bool ret = device::ascend::KernelPreBuildParallelCompile(kernel_graph_ptr);
+  return ret;
+}
+
 bool KernelBuild(const mindspore::session::KernelGraph *kernel_graph_ptr) {
   MS_EXCEPTION_IF_NULL(kernel_graph_ptr);
   TbeUtils::LoadCache();

mindspore/ccsrc/device/ascend/kernel_build_ascend.h

@@ -22,6 +22,10 @@
 namespace mindspore {
 namespace device {
 namespace ascend {
+/**
+ * @brief kernel pre build for ascend.
+ */
+bool KernelPreBuild(const mindspore::session::KernelGraph *kernel_graph_ptr);
 /**
  * @brief kernel build for ascend.
  */

mindspore/ccsrc/kernel/kernel_fusion.cc

@@ -102,7 +102,8 @@ std::map<int32_t, KernelModPtr> KernelFusion(const std::vector<FusionScopeInfo>
   while (!build_manger->IsAllTaskFinish()) {
     int task_id = -1;
     char *task_result = nullptr;
-    auto ret = build_manger->WaitOne(&task_id, &task_result);
+    char *pre_build_result = nullptr;
+    auto ret = build_manger->WaitOne(&task_id, &task_result, &pre_build_result);
     if (!ret) {
       MS_EXCEPTION(ArgumentError) << "Build Failed. wait one ret:" << ret << ", task id:" << task_id;
     }

mindspore/ccsrc/kernel/tbe/tbe_adapter.h

@@ -27,7 +27,7 @@
 //       the TBE back-end operator implementation difference
 namespace mindspore {
 namespace kernel {
-enum kCreaterType : int { SINGLE_BUILD = 0, PREBUILD, OP_SELECT_FORMAT, CHECK_SUPPORTED };
+enum kCreaterType : int { SINGLE_BUILD = 0, PREBUILD, OP_SELECT_FORMAT, CHECK_SUPPORTED, OP_PRE_COMPILE };
 namespace tbe {
 using FAttrsPass = void (*)(const AnfNodePtr &anf_node, const std::vector<std::shared_ptr<OpAttr>> &op_info_attrs,
                             nlohmann::json *attrs_json);

mindspore/ccsrc/kernel/tbe/tbe_kernel_parallel_build.cc

@@ -42,6 +42,40 @@ constexpr auto kStartCompileOp = "start_compile_op";
 constexpr auto kWaitOne = "wait_one";
 constexpr auto kResetTaskInfo = "reset_task_info";
 
+bool TbeOpParallelPreBuild(const std::vector<AnfNodePtr> &anf_nodes) {
+  auto build_manger = std::make_shared<ParallelBuildManager>();
+  MS_EXCEPTION_IF_NULL(build_manger);
+  for (const auto &anf_node : anf_nodes) {
+    // gen kernel json
+    nlohmann::json kernel_json;
+    TbeKernelJsonCreator creator(OP_PRE_COMPILE);
+    if (!creator.GenTbeSingleKernelJson(anf_node, &kernel_json)) {
+      MS_LOG(ERROR) << "GenTbeSingleKernelJson failed";
+      return false;
+    }
+    kernel_json["compile_type"] = "pre_build";
+    // op build
+    auto task_id = build_manger->StartCompileOp(kernel_json);
+    build_manger->SavePreTaskInfo(task_id, anf_node);
+  }
+  while (!build_manger->IsAllPreTaskFinish()) {
+    int task_id = -1;
+    char *task_result = nullptr;
+    char *pre_build_result = nullptr;
+    auto ret = build_manger->WaitOne(&task_id, &task_result, &pre_build_result);
+    if (!ret) {
+      MS_EXCEPTION(ArgumentError) << "Pre Build Failed. wait one ret:" << ret << ", task id:" << task_id;
+    }
+
+    if ((task_result != nullptr) && (strcmp(task_result, "Success") != 0)) {
+      MS_EXCEPTION(ArgumentError) << "task pre compile Failed, task id:" << task_id << ", cause:" << task_result;
+    }
+
+    build_manger->PreTaskFinishProcess(task_id, pre_build_result);
+  }
+  return true;
+}
+
 bool TbeOpParallelBuild(std::vector<AnfNodePtr> anf_nodes) {
   auto build_manger = std::make_shared<ParallelBuildManager>();
   MS_EXCEPTION_IF_NULL(build_manger);
@@ -82,7 +116,8 @@ bool TbeOpParallelBuild(std::vector<AnfNodePtr> anf_nodes) {
   while (!build_manger->IsAllTaskFinish()) {
     int task_id = -1;
     char *task_result = nullptr;
-    auto ret = build_manger->WaitOne(&task_id, &task_result);
+    char *pre_build_result = nullptr;
+    auto ret = build_manger->WaitOne(&task_id, &task_result, &pre_build_result);
     if (!ret) {
       MS_EXCEPTION(ArgumentError) << "Build Failed. wait one ret:" << ret << ", task id:" << task_id;
     }
@@ -116,7 +151,7 @@ int32_t ParallelBuildManager::StartCompileOp(const nlohmann::json &kernel_json)
   return task_id;
 }
 
-bool ParallelBuildManager::WaitOne(int *task_id, char **task_result) const {
+bool ParallelBuildManager::WaitOne(int *task_id, char **task_result, char **pre_build_result) const {
   MS_LOG(INFO) << "wait task start.";
   MS_EXCEPTION_IF_NULL(task_id);
   MS_EXCEPTION_IF_NULL(task_result);
@@ -128,10 +163,15 @@ bool ParallelBuildManager::WaitOne(int *task_id, char **task_result) const {
     MS_EXCEPTION(ArgumentError) << "Failed to call function wait_one";
     return false;
   }
-  (void)PyArg_ParseTuple(pRes, "is", task_id, task_result);
+  (void)PyArg_ParseTuple(pRes, "iss", task_id, task_result, pre_build_result);
   return true;
 }
 
+void ParallelBuildManager::SavePreTaskInfo(int32_t task_id, const mindspore::AnfNodePtr &anf_node) {
+  MS_LOG(INFO) << "SavePreTaskInfo, task id: " << task_id;
+  pre_task_map_[task_id] = anf_node;
+}
+
 void ParallelBuildManager::SaveTaskInfo(int32_t task_id, const mindspore::AnfNodePtr &anf_node,
                                         const std::string &json_name, const std::vector<size_t> &input_size_list,
                                         const std::vector<size_t> &output_size_list, int32_t scope_id) {
@@ -150,11 +190,24 @@ void ParallelBuildManager::SaveTaskInfo(int32_t task_id, const mindspore::AnfNod
   task_map_[task_id] = task_info;
 }
 
+bool ParallelBuildManager::IsAllPreTaskFinish() const {
+  MS_LOG(INFO) << "wait pre build process task_num: " << pre_task_map_.size();
+  return pre_task_map_.empty();
+}
+
 bool ParallelBuildManager::IsAllTaskFinish() const {
   MS_LOG(INFO) << "wait process task_num: " << task_map_.size();
   return task_map_.empty();
 }
 
+void ParallelBuildManager::PreTaskFinishProcess(int32_t task_id, const std::string &pre_build_result) {
+  auto task_iter = pre_task_map_.find(task_id);
+  if (task_iter == pre_task_map_.end()) {
+    MS_EXCEPTION(ArgumentError) << "can find pre task_id:" << task_id;
+  }
+  (void)pre_task_map_.erase(task_iter);
+}
+
 std::pair<int32_t, KernelModPtr> ParallelBuildManager::TaskFinishProcess(int32_t task_id, bool set_kernel_mod) {
   auto task_iter = task_map_.find(task_id);
   if (task_iter == task_map_.end()) {

mindspore/ccsrc/kernel/tbe/tbe_kernel_parallel_build.h

@@ -26,6 +26,7 @@
 #include <nlohmann/json.hpp>
 namespace mindspore {
 namespace kernel {
+bool TbeOpParallelPreBuild(const std::vector<AnfNodePtr> &anf_nodes);
 bool TbeOpParallelBuild(std::vector<AnfNodePtr> anf_nodes);
 
 struct KernelBuildTaskInfo {
@@ -42,6 +43,7 @@ class ParallelBuildManager {
   ParallelBuildManager();
   ~ParallelBuildManager();
   int32_t StartCompileOp(const nlohmann::json &kernel_json) const;
+  void SavePreTaskInfo(int32_t task_id, const AnfNodePtr &anf_node);
   void SaveTaskInfo(int32_t task_id, const AnfNodePtr &anf_node, const std::string &json_name,
                     const std::vector<size_t> &input_size_list, const std::vector<size_t> &output_size_list,
                     int32_t scope_id = 0);
@@ -52,8 +54,10 @@ class ParallelBuildManager {
                      const std::vector<size_t> &input_size_list, const std::vector<size_t> &output_size_list,
                      AnfNode *node) const;
 
-  bool WaitOne(int *task_id, char **task_result) const;
+  bool WaitOne(int *task_id, char **task_result, char **pre_build_result) const;
+  bool IsAllPreTaskFinish() const;
   bool IsAllTaskFinish() const;
+  void PreTaskFinishProcess(int32_t task_id, const std::string &pre_build_result);
   std::pair<int32_t, KernelModPtr> TaskFinishProcess(int32_t task_id, bool set_kernel_mod = true);
   KernelModPtr GenKernelMod(const string &json_name, const string &processor,
                             const std::vector<size_t> &input_size_list, const std::vector<size_t> &output_size_list,
@@ -62,6 +66,7 @@ class ParallelBuildManager {
 
  private:
   PyObject *tbe_parallel_compiler_;
+  std::map<int32_t, AnfNodePtr> pre_task_map_;
   std::map<int32_t, KernelBuildTaskInfo> task_map_;
   std::vector<KernelBuildTaskInfo> same_op_list_;
 };

mindspore/ccsrc/session/ascend_session.cc

@@ -550,6 +550,7 @@ void AscendSession::InitRuntimeResource() {
 
 void AscendSession::HardwareOptimize(const std::shared_ptr<KernelGraph> &kernel_graph) const {
   MS_LOG(INFO) << "HardwareOptimize start!";
+  device::ascend::KernelPreBuild(kernel_graph.get());
   opt::AscendBackendOptimization(kernel_graph);
   MS_EXCEPTION_IF_NULL(kernel_graph);
   kernel_graph->SetExecOrderByDefault();