diff --git a/paddle/fluid/distributed/collective/CMakeLists.txt b/paddle/fluid/distributed/collective/CMakeLists.txt
index de9059228a61bd06a60093595987a63248b089cd..7bc0d1310bf65402272009cda5cab7b71f1b24bf 100644
--- a/paddle/fluid/distributed/collective/CMakeLists.txt
+++ b/paddle/fluid/distributed/collective/CMakeLists.txt
@@ -1,5 +1,5 @@
 cc_library(
-  processgroup
+  process_group
   SRCS ProcessGroup.cc
   DEPS dense_tensor)
 cc_library(
@@ -9,20 +9,20 @@ cc_library(
 cc_library(
   eager_reducer
   SRCS reducer.cc
-  DEPS eager_api processgroup process_group_stream phi_api string_helper)
+  DEPS eager_api process_group process_group_stream phi_api string_helper)
 
 if(WITH_DISTRIBUTE)
   cc_library(
-    processgroup_gloo
-    SRCS ProcessGroupGloo.cc
+    process_group_gloo
+    SRCS process_group_gloo.cc
     DEPS phi_api eager_api gloo_wrapper)
 endif()
 
 if(WITH_NCCL OR WITH_RCCL)
   cc_library(
     process_group_nccl
-    SRCS process_group_nccl.cc nccl_tools.cc Common.cc check.cc
-    DEPS processgroup
+    SRCS process_group_nccl.cc nccl_tools.cc common.cc check.cc
+    DEPS process_group
          process_group_stream
          place
          enforce
@@ -34,23 +34,23 @@ endif()
 
 if(WITH_XPU_BKCL)
   cc_library(
-    processgroup_bkcl
-    SRCS ProcessGroupBKCL.cc BKCLTools.cc Common.cc
-    DEPS processgroup place enforce collective_helper device_context
+    process_group_bkcl
+    SRCS ProcessGroupBKCL.cc BKCLTools.cc common.cc
+    DEPS process_group place enforce collective_helper device_context
          dense_tensor)
 endif()
 
 if(WITH_MPI)
   cc_library(
-    processgroup_mpi
-    SRCS ProcessGroupMPI.cc MPITools.cc Common.cc
+    process_group_mpi
+    SRCS process_group_mpi.cc mpi_tools.cc common.cc
     DEPS collective_helper device_context)
 endif()
 
 if(WITH_CUSTOM_DEVICE)
   cc_library(
-    processgroup_custom
-    SRCS ProcessGroupCustom.cc CustomCCLTools.cc Common.cc
-    DEPS processgroup phi_backends place enforce collective_helper
+    process_group_custom
+    SRCS ProcessGroupCustom.cc CustomCCLTools.cc common.cc
+    DEPS process_group phi_backends place enforce collective_helper
          device_context)
 endif()
diff --git a/paddle/fluid/distributed/collective/HCCLTools.cc b/paddle/fluid/distributed/collective/HCCLTools.cc
deleted file mode 100644
index 7ab3aa87b08fd407cead05d12b1dd993c6447629..0000000000000000000000000000000000000000
--- a/paddle/fluid/distributed/collective/HCCLTools.cc
+++ /dev/null
@@ -1,48 +0,0 @@
-// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include "paddle/fluid/distributed/collective/HCCLTools.h"
-
-#include "paddle/fluid/distributed/collective/Types.h"
-
-namespace paddle {
-namespace distributed {
-
-HcclReduceOp ToHCCLRedType(ReduceOp reduction) {
-  static const std::map<ReduceOp, HcclReduceOp> red_type = {
-      {ReduceOp::MIN, HCCL_REDUCE_MIN},
-      {ReduceOp::MAX, HCCL_REDUCE_MAX},
-      {ReduceOp::SUM, HCCL_REDUCE_SUM},
-      {ReduceOp::PRODUCT, HCCL_REDUCE_PROD},
-  };
-  auto it = red_type.find(reduction);
-  PADDLE_ENFORCE_EQ(
-      it != red_type.end(),
-      true,
-      platform::errors::InvalidArgument("Invalid hccl reduction. "
-                                        "Must be Min | Max | Prod | Sum"));
-  return it->second;
-}
-
-std::string SerializeHCCLUniqueId(const HcclRootInfo& hcclID) {
-  const uint8_t* bytes = reinterpret_cast<const uint8_t*>(&hcclID);
-  std::ostringstream oss;
-  for (size_t i = 0; i < sizeof(hcclID); ++i) {
-    oss << std::hex << static_cast<int>(bytes[i]);
-  }
-  return oss.str();
-}
-
-}  //  namespace distributed
-}  //  namespace paddle
diff --git a/paddle/fluid/distributed/collective/HCCLTools.h b/paddle/fluid/distributed/collective/HCCLTools.h
deleted file mode 100644
index 89ce00fe87428ac6e449cab85e6919174dcbbed5..0000000000000000000000000000000000000000
--- a/paddle/fluid/distributed/collective/HCCLTools.h
+++ /dev/null
@@ -1,184 +0,0 @@
-// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#pragma once
-
-#include <error.h>
-
-#include <string>
-
-#include "paddle/fluid/distributed/collective/Types.h"
-#include "paddle/fluid/framework/data_type.h"
-#include "paddle/fluid/framework/variable.h"
-#include "paddle/fluid/platform/collective_helper.h"
-#include "paddle/fluid/platform/device/npu/enforce_npu.h"
-#include "paddle/fluid/platform/device/npu/npu_info.h"
-#include "paddle/fluid/platform/device_context.h"
-#include "paddle/fluid/platform/enforce.h"
-#include "paddle/utils/variant.h"
-
-namespace paddle {
-namespace distributed {
-
-class NPUEventManager {
- public:
-  NPUEventManager() = default;
-
-  ~NPUEventManager() {
-    if (is_created_) {
-      platform::NPUDeviceGuard guard(device_index_);
-      platform::NPUEventDestroy(event_);
-    }
-  }
-
-  NPUEventManager(const NPUEventManager&) = delete;
-  NPUEventManager& operator=(const NPUEventManager&) = delete;
-
-  NPUEventManager(NPUEventManager&& other) {
-    std::swap(is_created_, other.is_created_);
-    std::swap(device_index_, other.device_index_);
-    std::swap(event_, other.event_);
-  }
-
-  NPUEventManager& operator=(NPUEventManager&& other) {
-    std::swap(is_created_, other.is_created_);
-    std::swap(device_index_, other.device_index_);
-    std::swap(event_, other.event_);
-    return *this;
-  }
-
-  bool IsCreated() const { return is_created_; }
-  bool DeviceId() const { return device_index_; }
-  aclrtEvent GetRawNPUEvent() const { return event_; }
-
-  void Record(const paddle::platform::NPUDeviceContext& ctx) {
-    auto device_index = ctx.GetPlace().device;
-    if (!is_created_) {
-      CreateEvent(device_index);
-    }
-    PADDLE_ENFORCE_EQ(device_index,
-                      device_index_,
-                      platform::errors::PreconditionNotMet(
-                          "NPUDeviceContext's device %d does not match"
-                          "Event's device %d",
-                          device_index,
-                          device_index_));
-
-    platform::NPUDeviceGuard guard(device_index_);
-    platform::NPUEventRecord(event_, ctx.stream());
-  }
-
-  bool Query() const {
-    aclrtEventStatus status = ACL_EVENT_STATUS_COMPLETE;
-    platform::NPUEventQuery(event_, &status);
-    if (status == ACL_EVENT_STATUS_COMPLETE) {
-      return true;
-    }
-    return false;
-  }
-
-  void Block(const paddle::platform::NPUDeviceContext& ctx) const {
-    if (is_created_) {
-      auto device_index = ctx.GetPlace().device;
-      PADDLE_ENFORCE_EQ(device_index,
-                        device_index_,
-                        platform::errors::PreconditionNotMet(
-                            "phi::GPUContext's device %d does not match"
-                            "Event's device %d",
-                            device_index,
-                            device_index_));
-      platform::NPUDeviceGuard guard(device_index_);
-      platform::NPUStreamWaitEvent(ctx.stream(), event_);
-    }
-  }
-
- private:
-  bool is_created_{false};
-  aclrtEvent event_{};
-  int8_t device_index_{0};
-
- private:
-  void CreateEvent(int device_index) {
-    device_index_ = device_index;
-    platform::NPUDeviceGuard guard(device_index);
-    platform::NPUEventCreate(&event_);
-    is_created_ = true;
-  }
-};
-
-class HCCLCommManager {
- public:
-  explicit HCCLCommManager(HcclComm hcclComm) : hccl_comm_(hcclComm) {}
-
-  HCCLCommManager() : HCCLCommManager(nullptr) {}
-
-  ~HCCLCommManager() noexcept {
-    std::unique_lock<std::mutex> lock(mutex_);
-    if (hccl_comm_) {
-      platform::dynload::HcclCommDestroy(hccl_comm_);
-    }
-  }
-
-  static std::shared_ptr<HCCLCommManager> Create(int num_ranks,
-                                                 int rank,
-                                                 HcclRootInfo* comm_id,
-                                                 HcclComm hccl_comm) {
-    auto hccl_manager = std::make_shared<HCCLCommManager>();
-    auto ret = platform::dynload::HcclCommInitRootInfo(
-        num_ranks, comm_id, rank, &hccl_comm);
-    using __NPU_STATUS_TYPE__ = decltype(ret);
-    constexpr auto __success_type__ =
-        platform::details::NPUStatusType<__NPU_STATUS_TYPE__>::kSuccess;
-    if (UNLIKELY(ret != __success_type__)) {
-      VLOG(0) << "Error: create hccl_id error.";
-      exit(-1);
-    }
-
-    hccl_manager->hccl_id_ = comm_id;
-    hccl_manager->rank_ = rank;
-    hccl_manager->hccl_comm_ = hccl_comm;
-    return hccl_manager;
-  }
-
-  HcclRootInfo* GetHcclId() const {
-    std::unique_lock<std::mutex> lock(mutex_);
-    return hccl_id_;
-  }
-
-  HcclComm GetHcclComm() const {
-    std::unique_lock<std::mutex> lock(mutex_);
-    return hccl_comm_;
-  }
-
-  HCCLCommManager(const HCCLCommManager&) = delete;
-  HCCLCommManager& operator=(const HCCLCommManager&) = delete;
-  HCCLCommManager& operator=(HCCLCommManager&& other) = delete;
-
-  HCCLCommManager(HCCLCommManager&& other) {
-    std::unique_lock<std::mutex> lock(other.mutex_);
-    std::swap(hccl_comm_, other.hccl_comm_);
-  }
-
- protected:
-  HcclComm hccl_comm_;
-  HcclRootInfo* hccl_id_;
-  int rank_;
-  mutable std::mutex mutex_;
-};
-
-HcclReduceOp ToHCCLRedType(ReduceOp reduction);
-std::string SerializeHCCLUniqueId(const HcclRootInfo& hcclID);
-
-}  // namespace distributed
-}  // namespace paddle
diff --git a/paddle/fluid/distributed/collective/ProcessGroupBKCL.cc b/paddle/fluid/distributed/collective/ProcessGroupBKCL.cc
index 135621ff5f66ef88f94269a493cf47504c32c5f2..25a85f679a3b8f6f85ebf9c3e127053dc58419c6 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupBKCL.cc
+++ b/paddle/fluid/distributed/collective/ProcessGroupBKCL.cc
@@ -15,7 +15,7 @@
 #include "paddle/fluid/distributed/collective/ProcessGroupBKCL.h"
 
 #include "paddle/fluid/distributed/collective/BKCLTools.h"
-#include "paddle/fluid/distributed/collective/Common.h"
+#include "paddle/fluid/distributed/collective/common.h"
 #include "paddle/fluid/platform/device/xpu/bkcl_helper.h"
 #include "paddle/fluid/platform/device/xpu/xpu_info.h"
 #include "paddle/fluid/platform/device_context.h"
diff --git a/paddle/fluid/distributed/collective/ProcessGroupCustom.cc b/paddle/fluid/distributed/collective/ProcessGroupCustom.cc
index 7f75fc1784668832d2cabf25bcf501dd360f42aa..e4e79d8be623affad6387bf3b91d035637790689 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupCustom.cc
+++ b/paddle/fluid/distributed/collective/ProcessGroupCustom.cc
@@ -14,8 +14,8 @@
 
 #include "paddle/fluid/distributed/collective/ProcessGroupCustom.h"
 
-#include "paddle/fluid/distributed/collective/Common.h"
 #include "paddle/fluid/distributed/collective/CustomCCLTools.h"
+#include "paddle/fluid/distributed/collective/common.h"
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/place.h"
diff --git a/paddle/fluid/distributed/collective/Common.cc b/paddle/fluid/distributed/collective/common.cc
similarity index 97%
rename from paddle/fluid/distributed/collective/Common.cc
rename to paddle/fluid/distributed/collective/common.cc
index d5cac8ec687ad8ba46e65f68ecdb0f99b73d57a6..e60ecf9b8dcb5ae9fb81271eff148ae619ff8882 100644
--- a/paddle/fluid/distributed/collective/Common.cc
+++ b/paddle/fluid/distributed/collective/common.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/fluid/distributed/collective/Common.h"
+#include "paddle/fluid/distributed/collective/common.h"
 
 namespace paddle {
 namespace distributed {
diff --git a/paddle/fluid/distributed/collective/Common.h b/paddle/fluid/distributed/collective/common.h
similarity index 100%
rename from paddle/fluid/distributed/collective/Common.h
rename to paddle/fluid/distributed/collective/common.h
diff --git a/paddle/fluid/distributed/collective/MPITools.cc b/paddle/fluid/distributed/collective/mpi_tools.cc
similarity index 94%
rename from paddle/fluid/distributed/collective/MPITools.cc
rename to paddle/fluid/distributed/collective/mpi_tools.cc
index 042169728db4a3e7e5070e1f6b8066f361466f66..65831cf890cf68fd4422070d8e9e7977995d81a2 100644
--- a/paddle/fluid/distributed/collective/MPITools.cc
+++ b/paddle/fluid/distributed/collective/mpi_tools.cc
@@ -12,9 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/fluid/distributed/collective/MPITools.h"
-#include "paddle/fluid/distributed/collective/Common.h"
+#include "paddle/fluid/distributed/collective/mpi_tools.h"
 #include "paddle/fluid/distributed/collective/Types.h"
+#include "paddle/fluid/distributed/collective/common.h"
 
 namespace paddle {
 namespace distributed {
diff --git a/paddle/fluid/distributed/collective/MPITools.h b/paddle/fluid/distributed/collective/mpi_tools.h
similarity index 100%
rename from paddle/fluid/distributed/collective/MPITools.h
rename to paddle/fluid/distributed/collective/mpi_tools.h
diff --git a/paddle/fluid/distributed/collective/ProcessGroupGloo.cc b/paddle/fluid/distributed/collective/process_group_gloo.cc
similarity index 97%
rename from paddle/fluid/distributed/collective/ProcessGroupGloo.cc
rename to paddle/fluid/distributed/collective/process_group_gloo.cc
index 5e194b3ccf6063eb4da8a430ee4eb312e15e0133..d54775493853f0cce0fa26c6ff97deb1d73b4aae 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupGloo.cc
+++ b/paddle/fluid/distributed/collective/process_group_gloo.cc
@@ -28,8 +28,8 @@
 #include <gloo/reduce.h>
 #include <gloo/scatter.h>
 
-#include "paddle/fluid/distributed/collective/Common.h"
-#include "paddle/fluid/distributed/collective/ProcessGroupGloo.h"
+#include "paddle/fluid/distributed/collective/common.h"
+#include "paddle/fluid/distributed/collective/process_group_gloo.h"
 #include "paddle/fluid/framework/fleet/gloo_wrapper.h"
 #include "paddle/fluid/platform/enforce.h"
 
@@ -400,6 +400,15 @@ class AllgatherGlooTask : public ProcessGroupGloo::GlooTask {
   }
 };
 
+std::shared_ptr<ProcessGroup::Task> ProcessGroupGloo::AllGather(
+    phi::DenseTensor* out_tensor,
+    const phi::DenseTensor& in_tensor,
+    bool sync_op) {
+  std::vector<phi::DenseTensor> in_wrapper{in_tensor};
+  std::vector<phi::DenseTensor> out_wrapper{*out_tensor};
+  return AllGather(in_wrapper, out_wrapper, true);
+}
+
 std::shared_ptr<ProcessGroup::Task> ProcessGroupGloo::AllGather(
     phi::DenseTensor* out_tensor,
     const phi::DenseTensor& in_tensor,
diff --git a/paddle/fluid/distributed/collective/ProcessGroupGloo.h b/paddle/fluid/distributed/collective/process_group_gloo.h
similarity index 97%
rename from paddle/fluid/distributed/collective/ProcessGroupGloo.h
rename to paddle/fluid/distributed/collective/process_group_gloo.h
index 1beacfca4ffc6208a10dc5168be811adaf96e613..cdb5255c8db5f67121e5f86520d83abfe613bc60 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupGloo.h
+++ b/paddle/fluid/distributed/collective/process_group_gloo.h
@@ -120,6 +120,11 @@ class ProcessGroupGloo : public ProcessGroup {
       int64_t /*numel*/,   // for compatibility, no use now
       bool sync_op) override;
 
+  std::shared_ptr<ProcessGroup::Task> AllGather(
+      phi::DenseTensor* out_tensor,
+      const phi::DenseTensor& in_tensor,
+      bool sync_op) override;
+
   std::shared_ptr<ProcessGroup::Task> AllReduce(
       phi::DenseTensor* out_tensor,
       const phi::DenseTensor& in_tensor,
diff --git a/paddle/fluid/distributed/collective/ProcessGroupMPI.cc b/paddle/fluid/distributed/collective/process_group_mpi.cc
similarity index 99%
rename from paddle/fluid/distributed/collective/ProcessGroupMPI.cc
rename to paddle/fluid/distributed/collective/process_group_mpi.cc
index 66edb52b66d02443b35028ad8a253690c29b02b4..796e0bb692f2407cb584e7ae640a032cd523cded 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupMPI.cc
+++ b/paddle/fluid/distributed/collective/process_group_mpi.cc
@@ -12,9 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include "paddle/fluid/distributed/collective/ProcessGroupMPI.h"
+#include "paddle/fluid/distributed/collective/process_group_mpi.h"
 #include <chrono>
-#include "paddle/fluid/distributed/collective/Common.h"
+#include "paddle/fluid/distributed/collective/common.h"
 
 constexpr int64_t kWaitBlockTImeout = 10;
 namespace paddle {
diff --git a/paddle/fluid/distributed/collective/ProcessGroupMPI.h b/paddle/fluid/distributed/collective/process_group_mpi.h
similarity index 99%
rename from paddle/fluid/distributed/collective/ProcessGroupMPI.h
rename to paddle/fluid/distributed/collective/process_group_mpi.h
index d877f856dc1f78573eaae023c457d88c4cd73971..635d81d610309165959d46da3620882fb03088de 100644
--- a/paddle/fluid/distributed/collective/ProcessGroupMPI.h
+++ b/paddle/fluid/distributed/collective/process_group_mpi.h
@@ -30,7 +30,7 @@
 #include "paddle/fluid/platform/device_context.h"
 
 #if defined(PADDLE_WITH_MPI)
-#include "paddle/fluid/distributed/collective/MPITools.h"
+#include "paddle/fluid/distributed/collective/mpi_tools.h"
 #endif
 
 namespace paddle {
diff --git a/paddle/fluid/distributed/collective/process_group_nccl.cc b/paddle/fluid/distributed/collective/process_group_nccl.cc
index 0859708f92cf4be5337175421e42e1625f0df15f..1353ea719a8219761d2b874cd771aa442a9ceba0 100644
--- a/paddle/fluid/distributed/collective/process_group_nccl.cc
+++ b/paddle/fluid/distributed/collective/process_group_nccl.cc
@@ -14,8 +14,8 @@
 
 #include "paddle/fluid/distributed/collective/process_group_nccl.h"
 
-#include "paddle/fluid/distributed/collective/Common.h"
 #include "paddle/fluid/distributed/collective/check.h"
+#include "paddle/fluid/distributed/collective/common.h"
 #include "paddle/fluid/distributed/collective/nccl_tools.h"
 #include "paddle/fluid/distributed/collective/utils.h"
 #include "paddle/fluid/platform/device/gpu/nccl_helper.h"
diff --git a/paddle/fluid/pybind/CMakeLists.txt b/paddle/fluid/pybind/CMakeLists.txt
index 37e085b82bc08277d0c4bbafe0dc1336cf605bb3..55accf7e046b1cc30f6d8f025b2d26f72806e892 100755
--- a/paddle/fluid/pybind/CMakeLists.txt
+++ b/paddle/fluid/pybind/CMakeLists.txt
@@ -155,21 +155,21 @@ if(WITH_CUSTOM_DEVICE)
 endif()
 
 if(WITH_PYTHON)
-  set(PYBIND_DEPS ${PYBIND_DEPS} processgroup eager_reducer)
+  set(PYBIND_DEPS ${PYBIND_DEPS} process_group eager_reducer)
   if(WITH_NCCL OR WITH_RCCL)
     set(PYBIND_DEPS ${PYBIND_DEPS} process_group_nccl)
   endif()
   if(WITH_XPU_BKCL)
-    set(PYBIND_DEPS ${PYBIND_DEPS} processgroup_bkcl)
+    set(PYBIND_DEPS ${PYBIND_DEPS} process_group_bkcl)
   endif()
   if(WITH_GLOO)
-    set(PYBIND_DEPS ${PYBIND_DEPS} processgroup_gloo)
+    set(PYBIND_DEPS ${PYBIND_DEPS} process_group_gloo)
   endif()
   if(WITH_MPI)
-    set(PYBIND_DEPS ${PYBIND_DEPS} processgroup_mpi)
+    set(PYBIND_DEPS ${PYBIND_DEPS} process_group_mpi)
   endif()
   if(WITH_CUSTOM_DEVICE)
-    set(PYBIND_DEPS ${PYBIND_DEPS} processgroup_custom)
+    set(PYBIND_DEPS ${PYBIND_DEPS} process_group_custom)
   endif()
   if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
     set(DISTRIBUTE_COMPILE_FLAGS "${DISTRIBUTE_COMPILE_FLAGS} -faligned-new")
diff --git a/paddle/fluid/pybind/distributed_py.cc b/paddle/fluid/pybind/distributed_py.cc
index e9d59132d0e4605f5f160b53c0d94157a5708bae..913d8c22b9f694820ff27a24ef1f2ecf1707f110 100644
--- a/paddle/fluid/pybind/distributed_py.cc
+++ b/paddle/fluid/pybind/distributed_py.cc
@@ -38,7 +38,7 @@ limitations under the License. */
 #endif
 
 #if defined(PADDLE_WITH_MPI)
-#include "paddle/fluid/distributed/collective/ProcessGroupMPI.h"
+#include "paddle/fluid/distributed/collective/process_group_mpi.h"
 #endif
 
 #if defined(PADDLE_WITH_CUSTOM_DEVICE)
@@ -46,7 +46,7 @@ limitations under the License. */
 #endif
 
 #if defined(PADDLE_WITH_GLOO)
-#include "paddle/fluid/distributed/collective/ProcessGroupGloo.h"
+#include "paddle/fluid/distributed/collective/process_group_gloo.h"
 #include "paddle/fluid/distributed/store/tcp_store.h"
 #endif
 
diff --git a/paddle/phi/backends/CMakeLists.txt b/paddle/phi/backends/CMakeLists.txt
index c9e110ae7b8e40a17ac04d4771c770460f47106a..3ee28c3105803d7ab22bbb7a3b864cd69d8b204f 100644
--- a/paddle/phi/backends/CMakeLists.txt
+++ b/paddle/phi/backends/CMakeLists.txt
@@ -65,12 +65,12 @@ if(WITH_CUSTOM_DEVICE)
     DEPS phi_capi)
 endif()
 
-set(COMM_UTILS_DEPS processgroup)
+set(COMM_UTILS_DEPS process_group)
 if(WITH_NCCL OR WITH_RCCL)
   set(COMM_UTILS_DEPS ${PROCESS_GROUP_UTILS_DEPS} process_group_nccl)
 endif()
 if(WITH_CUSTOM_DEVICE)
-  set(COMM_UTILS_DEPS ${PROCESS_GROUP_UTILS_DEPS} processgroup_custom)
+  set(COMM_UTILS_DEPS ${PROCESS_GROUP_UTILS_DEPS} process_group_custom)
 endif()
 cc_library(
   processgroup_comm_utils
diff --git a/paddle/phi/kernels/CMakeLists.txt b/paddle/phi/kernels/CMakeLists.txt
index abe35f284d66407dae03305d77860625066f34cb..735ba7beaa2d1e78fa5d5351cef5f3e872fcb7f9 100644
--- a/paddle/phi/kernels/CMakeLists.txt
+++ b/paddle/phi/kernels/CMakeLists.txt
@@ -78,7 +78,7 @@ set(COMMON_KERNEL_DEPS
     gpc
     utf8proc)
 
-set(COMMON_KERNEL_DEPS ${COMMON_KERNEL_DEPS} processgroup)
+set(COMMON_KERNEL_DEPS ${COMMON_KERNEL_DEPS} process_group)
 if(WITH_NCCL OR WITH_RCCL)
   set(COMMON_KERNEL_DEPS ${COMMON_KERNEL_DEPS} process_group_nccl)
 endif()