[PIR] Fix bug in ParseDeviceContext (#60705)

* fix

* fix

* fix

* fix

* fix

Author: zhangbo9674

paddle/fluid/framework/new_executor/instruction/instruction_util.cc

@@ -91,13 +91,13 @@ platform::DeviceContext* ParseDeviceContext(
       return dev_ctx;
     }
 
-    if (op_name == interpreter::kMemcpyD2H) {
+    if (op_name.compare(paddle::dialect::MemcpyD2hOp::name()) == 0) {
       dev_ctx = ctx_manager.Get(std::string(kD2HStream), place, stream_priority)
                     .get()
                     .get();
       interpreter::SetDeviceCommContext(op, dev_ctx);
       return dev_ctx;
-    } else if (op_name == interpreter::kMemcpyH2D) {
+    } else if (op_name.compare(paddle::dialect::MemcpyH2dOp::name()) == 0) {
       dev_ctx = ctx_manager.Get(std::string(kH2DStream), place, stream_priority)
                     .get()
                     .get();
@@ -114,9 +114,11 @@ platform::DeviceContext* ParseDeviceContext(
     // DeviceContext passed from executor (see CAllReduceOpCUDAKernel in
     // c_allreduce_op.h). Now it is just a temporary solution for ONLY
     // c_allreduce_sum which is used in ResNet50 distributed training.
-    if (op_name == "c_allreduce_sum" && op_attributes.at("use_calc_stream")
-                                                .dyn_cast<pir::BoolAttribute>()
-                                                .data() == false) {
+    if ((op_name.compare(paddle::dialect::CAllreduceSumOp::name()) == 0 ||
+         op_name.compare(paddle::dialect::CAllreduceSum_Op::name()) == 0) &&
+        op_attributes.at("use_calc_stream")
+                .dyn_cast<pir::BoolAttribute>()
+                .data() == false) {
       int ring_id =
           op_attributes.at("ring_id").dyn_cast<pir::Int32Attribute>().data();
       if (FLAGS_dynamic_static_unified_comm) {

paddle/fluid/framework/new_executor/instruction/legacy_kernel_instruction.cc

@@ -90,7 +90,10 @@ LegacyKernelInstruction::LegacyKernelInstruction(
     }
     SetEventsToWaitInfo(events_to_wait);
   }
-  VLOG(6) << "finish process dist attributes";
+  VLOG(6) << "finish process dist attributes for " << op_name
+          << " : [execution_stream, stream_priority, scheduling_priority] = ["
+          << GetExecutionStream() << ", " << GetStreamPriority() << ", "
+          << GetSchedulingPriority() << "]";
 
   SetKernelType(AnalyseOpFuncType(op, place));
   VLOG(6) << "finish process analyse kernel type";
@@ -137,28 +140,29 @@ LegacyKernelInstruction::LegacyKernelInstruction(
 
   operator_base_ = BuildOperatorBase(op, *value_exec_info_, yaml_info_parser);
 
+  SetDeviceContext(
+      ParseDeviceContext(op,
+                         phi::DeviceContextPool::Instance().Get(
+                             phi::TransToPhiPlace(kernel_key.backend())),
+                         place,
+                         GetExecutionStream(),
+                         GetStreamPriority()));
+  VLOG(6) << "finish process device context";
+
   paddle::framework::VariableValueMap in_map;
   paddle::framework::VariableValueMap out_map;
-  auto dev_ctx = phi::DeviceContextPool::Instance().Get(
-      phi::TransToPhiPlace(kernel_key.backend()));
-
   runtime_context_ = std::make_shared<paddle::framework::RuntimeContext>(
       paddle::framework::RuntimeContext(in_map, out_map));
   BuildRuntimeContext(
       op, *value_exec_info, yaml_info_parser, runtime_context_.get());
 
-  kernel_context_ = new paddle::framework::ExecutionContext(
-      *operator_base_, *inner_scope, *dev_ctx, *(runtime_context_.get()));
+  kernel_context_ =
+      new paddle::framework::ExecutionContext(*operator_base_,
+                                              *inner_scope,
+                                              DeviceContext(),
+                                              *(runtime_context_.get()));
 
   VLOG(6) << "finish process kernel context";
-  SetDeviceContext(
-      ParseDeviceContext(op,
-                         phi::DeviceContextPool::Instance().Get(
-                             phi::TransToPhiPlace(kernel_key.backend())),
-                         place,
-                         GetExecutionStream(),
-                         GetStreamPriority()));
-  VLOG(6) << "finish process device context";
 
   InitInputsOutputsIds(op, *value_exec_info);
   VLOG(6) << "finish process inputs outputs index";

paddle/fluid/framework/new_executor/instruction/phi_kernel_instruction.cc

@@ -94,7 +94,10 @@ PhiKernelInstruction::PhiKernelInstruction(
     }
     SetEventsToWaitInfo(events_to_wait);
   }
-  VLOG(6) << "finish process dist attributes";
+  VLOG(6) << "finish process dist attributes for " << op_name
+          << " : [execution_stream, stream_priority, scheduling_priority] = ["
+          << GetExecutionStream() << ", " << GetStreamPriority() << ", "
+          << GetSchedulingPriority() << "]";
 
   SetKernelType(AnalyseOpFuncType(op, place));
   VLOG(6) << "finish process analyse kernel type";
@@ -137,6 +140,16 @@ PhiKernelInstruction::PhiKernelInstruction(
       phi_kernel_->IsValid(), true, "not found kernel for [%s]", kernel_name);
   VLOG(6) << "finish process select kernel";
 
+  platform::DeviceContext* dev_ctx =
+      ParseDeviceContext(op,
+                         phi::DeviceContextPool::Instance().Get(
+                             phi::TransToPhiPlace(kernel_key.backend())),
+                         place,
+                         GetExecutionStream(),
+                         GetStreamPriority());
+  SetDeviceContext(dev_ctx);
+  VLOG(6) << "finish process device context";
+
   BuildPhiContext<phi::KernelContext,
                   const phi::TensorBase*,
                   phi::TensorBase*,
@@ -145,19 +158,9 @@ PhiKernelInstruction::PhiKernelInstruction(
                   true>(
       op, *value_exec_info_, yaml_info_parser, &kernel_context_);
 
-  kernel_context_.SetDeviceContext(phi::DeviceContextPool::Instance().Get(
-      phi::TransToPhiPlace(kernel_key.backend())));
+  kernel_context_.SetDeviceContext(dev_ctx);
   VLOG(6) << "finish process kernel context";
 
-  SetDeviceContext(
-      ParseDeviceContext(op,
-                         phi::DeviceContextPool::Instance().Get(
-                             phi::TransToPhiPlace(kernel_key.backend())),
-                         place,
-                         GetExecutionStream(),
-                         GetStreamPriority()));
-  VLOG(6) << "finish process device context";
-
   InitInputsOutputsIds(op, *value_exec_info);
   VLOG(6) << "finish process inputs outputs index";