[new-exec] add completion_nofifier (#38447)

* add completion_nofifier * fix bug * unregist event waiter
PaddlePaddle · Dec 28, 2021 · 404a4a6 · 404a4a6
1 parent 1db61c3
commit 404a4a6
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 11 deletions.
diff --git a/paddle/fluid/framework/new_executor/interpretercore.cc b/paddle/fluid/framework/new_executor/interpretercore.cc
@@ -30,6 +30,7 @@ DECLARE_bool(check_nan_inf);
 DECLARE_bool(benchmark);
 
 constexpr const char* kExceptionCaught = "ExceptionCaught";
+constexpr const char* kTaskCompletion = "TaskCompletion";
 
 namespace paddle {
 namespace framework {
@@ -49,6 +50,7 @@ InterpreterCore::InterpreterCore(const platform::Place& place,
   gc_.reset(new InterpreterCoreGarbageCollector());
 
   exception_notifier_ = main_thread_blocker_.RegisterEvent(kExceptionCaught);
+  completion_notifier_ = main_thread_blocker_.RegisterEvent(kTaskCompletion);
 
   create_local_scope_ = FLAGS_new_executor_use_local_scope;
   if (FLAGS_new_executor_use_local_scope) {
@@ -69,6 +71,9 @@ InterpreterCore::~InterpreterCore() {
   // cancle gc's thread
   gc_.reset(nullptr);
 
+  exception_notifier_->UnregisterEvent();
+  completion_notifier_->UnregisterEvent();
+
   async_work_queue_.reset(nullptr);
 }
 
@@ -417,7 +422,7 @@ void InterpreterCore::ExecuteInstructionList(
     const std::vector<Instruction>& vec_instr) {
   async_work_queue_->PrepareAtomicDeps(dependecy_count_);
   async_work_queue_->PrepareAtomicVarRef(global_scope_->VecMetaInfo());
-  op_run_number_ = 0;
+  unfinished_op_numer_ = vec_instr.size();
 
   exception_holder_.Clear();
 
@@ -436,12 +441,6 @@ void InterpreterCore::ExecuteInstructionList(
     async_work_queue_->Cancel();
     exception_holder_.ReThrow();
   }
-
-  PADDLE_ENFORCE_EQ(
-      op_run_number_.load(), vec_instr.size(),
-      platform::errors::Fatal(
-          "Required op_run_number == %d, but received op_run_number = %d.",
-          vec_instr.size(), op_run_number_.load()));
 }
 
 void InterpreterCore::RunNextInstructions(
@@ -539,8 +538,15 @@ void InterpreterCore::RunInstructionAsync(size_t instr_id) {
       return;
     }
 
+    VLOG(4) << "unfinished_op_numer_: " << unfinished_op_numer_;
+    if (UNLIKELY(unfinished_op_numer_.fetch_sub(1, std::memory_order_relaxed) ==
+                 1)) {
+      if (completion_notifier_ != nullptr) {
+        completion_notifier_->NotifyEvent();
+      }
+    }
+
     interpreter::RecordEvent(instr_node, place_);
-    op_run_number_.fetch_add(1, std::memory_order_relaxed);
 
     RunNextInstructions(instr_node, &ready_ops);
   }

diff --git a/paddle/fluid/framework/new_executor/interpretercore.h b/paddle/fluid/framework/new_executor/interpretercore.h
@@ -101,14 +101,15 @@ class InterpreterCore {
   std::vector<Instruction> vec_instruction_;  // deconstruct before OpFuncNode
 
   std::vector<size_t> dependecy_count_;
-  std::atomic<size_t> op_run_number_{0};
+  std::atomic<size_t> unfinished_op_numer_{0};
   std::vector<std::vector<size_t>> input_var2op_info_;
 
   StreamAnalyzer stream_analyzer_;
   EventsWaiter main_thread_blocker_;
   std::unique_ptr<interpreter::AsyncWorkQueue> async_work_queue_;
   details::ExceptionHolder exception_holder_;
   std::shared_ptr<EventsWaiter::EventNotifier> exception_notifier_{nullptr};
+  std::shared_ptr<EventsWaiter::EventNotifier> completion_notifier_{nullptr};
 
   std::unique_ptr<InterpreterCoreGarbageCollector> gc_;
   std::vector<paddle::platform::DeviceEvent> gc_event_;

diff --git a/paddle/fluid/framework/new_executor/interpretercore_util.h b/paddle/fluid/framework/new_executor/interpretercore_util.h
@@ -60,13 +60,13 @@ class AsyncWorkQueue {
     // for execute host Kernel
     group_options.emplace_back(/*num_threads*/ host_num_threads,
                                /*allow_spinning*/ true,
-                               /*track_task*/ true,
+                               /*track_task*/ false,
                                /*detached*/ true,
                                /*events_waiter*/ waiter);
     // for launch device Kernel
     group_options.emplace_back(/*num_threads*/ 1,
                                /*allow_spinning*/ true,
-                               /*track_task*/ true,
+                               /*track_task*/ false,
                                /*detached*/ true,
                                /*events_waiter*/ waiter);
     queue_group_ = CreateWorkQueueGroup(group_options);