diff --git a/paddle/fluid/distributed/collective/process_group_nccl.cc b/paddle/fluid/distributed/collective/process_group_nccl.cc
index 7ffe00b8cd824..89f5dcb222e63 100644
--- a/paddle/fluid/distributed/collective/process_group_nccl.cc
+++ b/paddle/fluid/distributed/collective/process_group_nccl.cc
@@ -44,7 +44,7 @@ ProcessGroupNCCL::NCCLTask::NCCLTask(const Place& place,
                                      bool sync_op,
                                      bool use_calc_stream)
     : TaskStream(rank, comm_type, sync_op, use_calc_stream),
-      comm_event_(place),
+      comm_event_(place, platform::GenerateDeviceEventFlag()),
       task_place_(place) {}
 
 ProcessGroupNCCL::NCCLTask::~NCCLTask() = default;
@@ -506,7 +506,9 @@ void ProcessGroupNCCL::CreateNCCLEnvCache(const Place& place,
   auto nccl_comm_ctx = this->GetCommContext();
   comm_ctx->set_nccl_comm(nccl_comm_ctx->GetNcclComm());
 
-  place_to_calc_event_.emplace(place_key, place);
+  place_to_calc_event_.emplace(
+      place_key,
+      platform::DeviceEvent(place, platform::GenerateDeviceEventFlag()));
   place_to_calc_ctx_.emplace(place_key, calc_ctx);
   place_to_comm_ctx_.emplace(place_key, std::move(comm_ctx));
 
@@ -592,7 +594,7 @@ ProcessGroupNCCL::NCCLTask::NCCLTask(
     CommType CommType,
     const std::vector<phi::DenseTensor>& inputs)
     : TaskStream(rank, inputs, CommType),
-      comm_event_(places[0]),
+      comm_event_(places[0], platform::GenerateDeviceEventFlag()),
       task_place_(places[0]) {}
 
 // create NCCLManager cache for places_key
@@ -636,7 +638,9 @@ void ProcessGroupNCCL::CreateNCCLManagerCache(
   GroupEnd();
 
   // TODO(sunyilun): for compatibility, will be removed later
-  place_to_calc_event_.emplace(places_key, places[0]);
+  place_to_calc_event_.emplace(
+      places_key,
+      platform::DeviceEvent(places[0], platform::GenerateDeviceEventFlag()));
   place_to_calc_ctx_.emplace(
       places_key,
       static_cast<phi::GPUContext*>(
diff --git a/paddle/fluid/platform/device_event_base.h b/paddle/fluid/platform/device_event_base.h
index e2de1e5a9abe3..03fd7d4bb13f0 100644
--- a/paddle/fluid/platform/device_event_base.h
+++ b/paddle/fluid/platform/device_event_base.h
@@ -55,7 +55,7 @@ enum EventStatus {
 
 class DeviceEvent {
  public:
-  explicit DeviceEvent(const platform::Place& place, unsigned int flag = 0)
+  explicit DeviceEvent(const platform::Place& place, unsigned int flag)
       : event_(), place_(place), flag_(flag) {
     type_id_ = DeviceTypeToId(platform::Place2DeviceType(place));
     PADDLE_ENFORCE_LT(type_id_,
diff --git a/paddle/fluid/platform/device_event_test.cc b/paddle/fluid/platform/device_event_test.cc
index 7dfacc66437ae..072fa35b8749c 100644
--- a/paddle/fluid/platform/device_event_test.cc
+++ b/paddle/fluid/platform/device_event_test.cc
@@ -37,7 +37,7 @@ TEST(DeviceEvent, CUDA) {
 
   ASSERT_NE(context, nullptr);
   // case 1. test for event_creator
-  DeviceEvent event(place);
+  DeviceEvent event(place, paddle::platform::GenerateDeviceEventFlag());
   ASSERT_NE(event.GetEvent().get(), nullptr);
   bool status = event.Query();
   ASSERT_EQ(status, true);
@@ -86,7 +86,7 @@ TEST(DeviceEvent, CUDA) {
 
   ASSERT_NE(context, nullptr);
   // case 1. test for event_creator
-  DeviceEvent event(place);
+  DeviceEvent event(place, paddle::platform::GenerateDeviceEventFlag());
   ASSERT_NE(event.GetEvent().get(), nullptr);
   bool status = event.Query();
   ASSERT_EQ(status, true);
@@ -127,7 +127,7 @@ TEST(DeviceEvent, CUDA) {
 TEST(DeviceEvent, CPU) {
   using paddle::platform::CPUPlace;
   auto place = CPUPlace();
-  DeviceEvent event(place);
+  DeviceEvent event(place, paddle::platform::GenerateDeviceEventFlag());
   auto& pool = DeviceContextPool::Instance();
   auto* context = pool.Get(place);