Update record interface using part2 (#39694)

* fix RecordEvent interface

* modify default level to 4

* update interface use

* add const default trace level

* update record event interface using

* update record event interface using

* update operator.cc

* update part2

* update part1

* fix include profiler.h header in ps server

* fix include profiler.h header in ps server

* fix profiler.h header

paddle/fluid/framework/details/scope_buffered_monitor.cc

@@ -13,7 +13,7 @@
 // limitations under the License.
 
 #include "paddle/fluid/framework/details/scope_buffered_monitor.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 namespace paddle {
 namespace framework {
@@ -91,7 +91,8 @@ void ScopeBufferedMonitor::Apply(const std::function<void()> &callback,
                                  bool has_fetch) {
   std::unique_ptr<platform::RecordEvent> pre_local_exec_scopes_event(
       new platform::RecordEvent(
-          "ScopeBufferedMonitor::pre_local_exec_scopes_process"));
+          "ScopeBufferedMonitor::pre_local_exec_scopes_process",
+          platform::TracerEventType::UserDefined, 2));
   for (size_t scope_id = 0; scope_id < local_exec_scopes_.size(); ++scope_id) {
     pre_local_exec_scopes_.at(scope_id).clear();
     auto scopes = local_exec_scopes_.at(scope_id)->kids();
@@ -105,7 +106,8 @@ void ScopeBufferedMonitor::Apply(const std::function<void()> &callback,
 
   std::unique_ptr<platform::RecordEvent> post_local_exec_scopes_event(
       new platform::RecordEvent(
-          "ScopeBufferedMonitor::post_local_exec_scopes_process"));
+          "ScopeBufferedMonitor::post_local_exec_scopes_process",
+          platform::TracerEventType::UserDefined, 2));
   for (size_t scope_id = 0; scope_id < local_exec_scopes_.size(); ++scope_id) {
     post_local_exec_scopes_.at(scope_id).clear();
     auto scopes = local_exec_scopes_.at(scope_id)->kids();

paddle/fluid/framework/details/scope_buffered_ssa_graph_executor.cc

@@ -23,7 +23,7 @@
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/variable_helper.h"
 #include "paddle/fluid/platform/cuda_graph_with_memory_pool.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 namespace paddle {
 namespace framework {
@@ -75,7 +75,8 @@ FetchResultType ScopeBufferedSSAGraphExecutor::Run(
 #endif
 
   if (drop_scope_counter_ == 0) {
-    platform::RecordEvent e("InitLocalVars");
+    platform::RecordEvent e("InitLocalVars",
+                            platform::TracerEventType::UserDefined, 2);
     InitVariables();
   }
 
@@ -164,7 +165,8 @@ void ScopeBufferedSSAGraphExecutor::InitVariables() {
 }
 
 void ScopeBufferedSSAGraphExecutor::DropLocalExeScopes(bool need_wait) {
-  platform::RecordEvent drop_scope_event("DropLocalExeScopes");
+  platform::RecordEvent drop_scope_event(
+      "DropLocalExeScopes", platform::TracerEventType::UserDefined, 2);
   drop_scope_counter_ = 0;
   if (need_wait) {
     for (auto &p : places_) {

paddle/fluid/framework/details/sparse_all_reduce_op_handle.cc

@@ -25,7 +25,7 @@
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/platform/cuda_device_guard.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 DECLARE_bool(sync_nccl_allreduce);
 
@@ -66,7 +66,8 @@ SparseAllReduceOpHandle::SparseAllReduceOpHandle(
 }
 
 void SparseAllReduceOpHandle::RunImplEncoded() {
-  platform::RecordEvent record_event(Name());
+  platform::RecordEvent record_event(Name(),
+                                     platform::TracerEventType::UserDefined, 2);
 
   auto in_var_handles = DynamicCast<VarHandle>(this->Inputs());
   auto out_var_handles = DynamicCast<VarHandle>(this->Outputs());
@@ -279,6 +280,8 @@ bool SparseAllReduceOpHandle::IsEncoded() {
 }
 
 void SparseAllReduceOpHandle::RunImpl() {
+  platform::RecordEvent record_event(
+      Name(), platform::TracerEventType::Communication, 1);
   if (!IsEncoded()) {
     AllReduceOpHandle::RunImpl();
     return;

paddle/fluid/framework/details/threaded_ssa_graph_executor.cc

@@ -15,7 +15,7 @@
 #include "paddle/fluid/framework/details/threaded_ssa_graph_executor.h"
 
 #include "paddle/fluid/framework/ir/graph_helper.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 #if defined PADDLE_WITH_PSCORE
 #include "paddle/fluid/distributed/ps/service/communicator/communicator.h"
@@ -56,7 +56,8 @@ ThreadedSSAGraphExecutor::ThreadedSSAGraphExecutor(
 inline FetchResultType ThreadedSSAGraphExecutor::RunImpl(
     const std::vector<std::string> &fetch_tensors, bool return_merged) {
   std::unique_ptr<platform::RecordEvent> event(
-      new platform::RecordEvent("ThreadedSSAGraphExecutorPrepare"));
+      new platform::RecordEvent("ThreadedSSAGraphExecutorPrepare",
+                                platform::TracerEventType::UserDefined, 2));
   std::unique_ptr<OpDependentData> op_deps = op_deps_futures_.get();
   CopyOpDeps();
 

paddle/fluid/framework/executor.cc

@@ -22,6 +22,7 @@ limitations under the License. */
 #include "paddle/fluid/operators/controlflow/while_op_helper.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 #ifdef PADDLE_WITH_MKLDNN
 #include "paddle/fluid/platform/mkldnn_helper.h"
 #endif

paddle/fluid/framework/ir/cost_model.h

@@ -26,6 +26,7 @@
 #include "paddle/fluid/framework/ir/node.h"
 #include "paddle/fluid/framework/program_desc.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 #include "paddle/fluid/platform/variant.h"
 
 namespace paddle {

paddle/fluid/framework/new_executor/interpretercore.cc

@@ -388,7 +388,9 @@ void InterpreterCore::RunInstruction(const Instruction& instr_node) {
                            : global_scope_->GetMutableScope();
   auto op_with_kernel = dynamic_cast<const framework::OperatorWithKernel*>(op);
   {
-    platform::RecordEvent infershape_event("InferShape");
+    platform::RecordEvent infershape_event(
+        "InferShape", platform::TracerEventType::OperatorInner, 1,
+        platform::EventRole::kInnerOp);
     // If it is OperatorBase, InferShape do nothing.
     if (op_with_kernel != nullptr)
       op_with_kernel->Info().infer_shape_(
@@ -408,7 +410,9 @@ void InterpreterCore::RunInstruction(const Instruction& instr_node) {
     }
   }
   {
-    platform::RecordEvent compute_event("Compute");
+    platform::RecordEvent compute_event(
+        "Compute", platform::TracerEventType::OperatorInner, 1,
+        platform::EventRole::kInnerOp);
     if (op_with_kernel == nullptr) {
       instr_node.OpBase()->Run(*local_scope, place_);
     } else {

paddle/fluid/framework/new_executor/workqueue/nonblocking_threadpool.h

@@ -408,7 +408,8 @@ class ThreadPoolTempl {
       ec_.Notify(true);
       return false;
     }
-    platform::RecordEvent("SleepWaitForWork");
+    platform::RecordEvent("SleepWaitForWork",
+                          platform::TracerEventType::UserDefined, 2);
     ec_.CommitWait(waiter);
     blocked_--;
     return true;

paddle/fluid/framework/parallel_executor.cc

@@ -39,6 +39,7 @@ limitations under the License. */
 #include "paddle/fluid/platform/cuda_graph_with_memory_pool.h"
 #include "paddle/fluid/platform/event.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
 #include "paddle/fluid/platform/cuda_device_guard.h"

paddle/fluid/framework/tensor_util.cc

@@ -23,7 +23,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/data_type.h"
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/platform/complex.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 #include "paddle/phi/core/dense_tensor.h"
 

paddle/fluid/imperative/basic_engine.cc

@@ -410,7 +410,8 @@ void BasicEngine::Execute() {
     auto& inplace_grad_name_map = shared_cur_node->InplaceGradNameMap();
 
     for (auto& cur_op : *shared_cur_node) {
-      platform::RecordEvent op_type_record_event(cur_op.Type());
+      platform::RecordEvent op_type_record_event(
+          cur_op.Type(), platform::TracerEventType::Operator, 1);
 
       ++op_num;
 

paddle/fluid/imperative/layer.cc

@@ -233,7 +233,8 @@ void VarBase::ClearGradient(bool set_to_zero) {
         grad_t->mutable_value()->clear();
       }
     } else {
-      platform::RecordEvent record_event("ClearGradient");
+      platform::RecordEvent record_event(
+          "ClearGradient", platform::TracerEventType::UserDefined, 2);
       auto* grad_t =
           grad_var_->MutableVar()->GetMutable<framework::LoDTensor>();
       if (grad_t->IsInitialized()) {

paddle/fluid/imperative/partial_grad_engine.cc

@@ -30,7 +30,7 @@
 #include "paddle/fluid/imperative/op_base.h"
 #include "paddle/fluid/imperative/tracer.h"
 #include "paddle/fluid/platform/device_context.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 #include "paddle/fluid/string/string_helper.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 

paddle/fluid/imperative/tracer.cc

@@ -22,6 +22,7 @@
 #include "paddle/fluid/platform/denormal.h"
 #include "paddle/fluid/platform/device/device_wrapper.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 #include "paddle/fluid/string/string_helper.h"
 
 DECLARE_bool(use_mkldnn);
@@ -171,7 +172,8 @@ void Tracer::TraceOp(const std::string& type, const NameVarMap<VarType>& ins,
                      const std::map<std::string, std::string>& inplace_map,
                      paddle::framework::AttributeMap* passed_default_attrs_,
                      bool use_default_attr_map) {
-  platform::RecordEvent op_type_record_event(type);
+  platform::RecordEvent op_type_record_event(
+      type, platform::TracerEventType::Operator, 2);
   platform::ScopedFlushDenormal flush;
   VLOG(1) << "Trace Op: " << type;
   if (FLAGS_use_mkldnn) {

paddle/fluid/inference/tensorrt/plugin/trt_plugin.h

@@ -24,7 +24,7 @@
 #include "paddle/fluid/inference/tensorrt/helper.h"
 #include "paddle/fluid/inference/tensorrt/plugin/trt_plugin_utils.h"
 #include "paddle/fluid/platform/enforce.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 namespace nvinfer1 {
 class ITensor;

paddle/fluid/inference/tests/api/tester_helper.h

@@ -38,7 +38,7 @@
 #include "paddle/fluid/inference/tests/api/config_printer.h"
 #include "paddle/fluid/inference/tests/test_helper.h"
 #include "paddle/fluid/inference/utils/benchmark.h"
-#include "paddle/fluid/platform/profiler.h"
+#include "paddle/fluid/platform/profiler/event_tracing.h"
 
 DEFINE_string(model_name, "", "model name");
 DEFINE_string(infer_model, "", "model path");