refine allocation cmake (#44241)

* build into one static library

* move memory/detail to memory/allocation

* fix bug

* fix profiler

* fix framework_proto

* fix deps

* fix inference compilation

* fix rocm compile

* follow comments

* fix buddy_allocator_test

cmake/inference_lib.cmake

@@ -427,10 +427,8 @@ copy(
 set(module "memory")
 copy(
   fluid_lib_dist
-  SRCS ${src_dir}/${module}/*.h ${src_dir}/${module}/detail/*.h
-       ${src_dir}/${module}/allocation/*.h
-  DSTS ${dst_dir}/${module} ${dst_dir}/${module}/detail
-       ${dst_dir}/${module}/allocation)
+  SRCS ${src_dir}/${module}/allocation/*.h
+  DSTS ${dst_dir}/${module}/allocation)
 
 set(module "platform")
 set(platform_lib_deps profiler_proto errors)

paddle/fluid/memory/CMakeLists.txt

-add_subdirectory(detail)
 add_subdirectory(allocation)
 
 if(WITH_MKLDNN)
@@ -10,7 +9,7 @@ endif()
 cc_library(
   malloc
   SRCS malloc.cc
-  DEPS place enforce allocator_facade profiler ${MKLDNN_CTX_DEPS})
+  DEPS place enforce allocator profiler ${MKLDNN_CTX_DEPS})
 cc_library(
   memcpy
   SRCS memcpy.cc

paddle/fluid/memory/allocation/CMakeLists.txt

-cc_library(
-  allocator
-  SRCS allocator.cc
-  DEPS place stats profiler)
-cc_library(
-  cpu_allocator
-  SRCS cpu_allocator.cc
-  DEPS allocator)
-cc_library(
-  locked_allocator
-  SRCS locked_allocator.cc
-  DEPS allocator)
-cc_library(
-  buffered_allocator
-  SRCS buffered_allocator.cc
-  DEPS allocator)
-cc_library(
-  best_fit_allocator
-  SRCS best_fit_allocator.cc
-  DEPS allocator)
-cc_library(
-  naive_best_fit_allocator
-  SRCS naive_best_fit_allocator.cc
-  DEPS allocator buddy_allocator)
-cc_test(
-  naive_best_fit_allocator_test
-  SRCS naive_best_fit_allocator_test.cc
-  DEPS naive_best_fit_allocator)
-cc_test(
-  buffered_allocator_test
-  SRCS buffered_allocator_test.cc
-  DEPS locked_allocator buffered_allocator cpu_allocator best_fit_allocator)
+include(ExternalProject)
+
+set(ALLOCATOR_DEPS place stats profiler)
+set(ALLOCATOR_SRCS
+    allocator.cc
+    cpu_allocator.cc
+    locked_allocator.cc
+    aligned_allocator.cc
+    buffered_allocator.cc
+    best_fit_allocator.cc
+    naive_best_fit_allocator.cc
+    allocator_strategy.cc
+    allocator_facade.cc
+    auto_growth_best_fit_allocator.cc
+    virtual_memory_auto_growth_best_fit_allocator.cc
+    retry_allocator.cc
+    memory_block.cc
+    memory_block_desc.cc
+    meta_cache.cc
+    buddy_allocator.cc
+    system_allocator.cc)
 
-if(WITH_MKLDNN)
-  set(MKLDNN_CTX_DEPS mkldnn)
-else()
-  set(MKLDNN_CTX_DEPS)
+if(WITH_GPU OR WITH_ROCM)
+  list(
+    APPEND
+    ALLOCATOR_SRCS
+    cuda_allocator.cc
+    cuda_managed_allocator.cc
+    pinned_allocator.cc
+    stream_safe_cuda_allocator.cc
+    thread_local_allocator.cc)
+  list(APPEND ALLOCATOR_DEPS cuda_device_guard gpu_info dynload_cuda)
 endif()
 
 if(WITH_GPU)
-  nv_library(
-    cuda_allocator
-    SRCS cuda_allocator.cc
-    DEPS allocator cuda_device_guard stats)
-  nv_library(
-    cuda_managed_allocator
-    SRCS cuda_managed_allocator.cc
-    DEPS allocator cuda_device_guard gpu_info)
-  nv_library(
-    pinned_allocator
-    SRCS pinned_allocator.cc
-    DEPS allocator)
-  nv_library(
-    stream_safe_cuda_allocator
-    SRCS stream_safe_cuda_allocator.cc
-    DEPS allocator cuda_graph)
-  nv_library(
-    thread_local_allocator
-    SRCS thread_local_allocator.cc
-    DEPS allocator)
+  list(APPEND ALLOCATOR_DEPS cuda_graph)
+endif()
 
-  cc_test(
-    thread_local_allocator_test
-    SRCS thread_local_allocator_test.cc
-    DEPS thread_local_allocator)
-  if(CUDA_VERSION GREATER_EQUAL 10.2)
-    nv_library(
-      cuda_virtual_mem_allocator
-      SRCS cuda_virtual_mem_allocator.cc
-      DEPS dynload_cuda)
+if(CUDA_VERSION VERSION_GREATER_EQUAL 10.2)
+  list(APPEND ALLOCATOR_SRCS cuda_virtual_mem_allocator.cc)
+endif()
+
+if(NOT WIN32)
+  list(APPEND ALLOCATOR_SRCS mmap_allocator.cc)
+  if(WITH_GPU)
+    list(APPEND ALLOCATOR_SRCS cuda_ipc_allocator.cc)
   endif()
 endif()
 
-if(WITH_ROCM)
-  hip_library(
-    cuda_allocator
-    SRCS cuda_allocator.cc
-    DEPS allocator cuda_device_guard stats)
-  hip_library(
-    cuda_managed_allocator
-    SRCS cuda_managed_allocator.cc
-    DEPS allocator cuda_device_guard gpu_info)
-  hip_library(
-    pinned_allocator
-    SRCS pinned_allocator.cc
-    DEPS allocator)
-  hip_library(
-    stream_safe_cuda_allocator
-    SRCS stream_safe_cuda_allocator.cc
-    DEPS allocator)
-  hip_library(
-    thread_local_allocator
-    SRCS thread_local_allocator.cc
-    DEPS allocator)
+if(WITH_ASCEND_CL)
+  list(APPEND ALLOCATOR_SRCS npu_allocator.cc npu_pinned_allocator.cc)
+  list(APPEND ALLOCATOR_DEPS npu_info)
+endif()
 
-  cc_test(
-    thread_local_allocator_test
-    SRCS thread_local_allocator_test.cc
-    DEPS thread_local_allocator)
+if(WITH_CUSTOM_DEVICE)
+  list(APPEND ALLOCATOR_SRCS custom_allocator.cc)
+  list(APPEND ALLOCATOR_DEPS device_manager)
 endif()
 
-if(WITH_ASCEND_CL)
-  cc_library(
-    npu_allocator
-    SRCS npu_allocator.cc
-    DEPS allocator npu_info)
-  cc_library(
-    npu_pinned_allocator
-    SRCS npu_pinned_allocator.cc
-    DEPS allocator npu_info)
+if(WITH_XPU)
+  list(APPEND ALLOCATOR_DEPS xpu_info)
+endif()
+
+if(WITH_IPU)
+  list(APPEND ALLOCATOR_DEPS ipu_info)
 endif()
 
-cc_library(
-  retry_allocator
-  SRCS retry_allocator.cc
+add_library(allocator "${ALLOCATOR_SRCS}")
+target_link_libraries(allocator ${ALLOCATOR_DEPS})
+# note: why only add dependency for framework_proto.
+# Because it is needed to generate framework.pb.h used in some header files.
+add_dependencies(allocator framework_proto)
+set_property(GLOBAL PROPERTY FLUID_MODULES allocator)
+
+cc_test(
+  naive_best_fit_allocator_test
+  SRCS naive_best_fit_allocator_test.cc
+  DEPS allocator)
+cc_test(
+  buffered_allocator_test
+  SRCS buffered_allocator_test.cc
   DEPS allocator)
 
-if(WITH_GPU OR WITH_ROCM)
-  set(AllocatorFacadeDeps
-      gpu_info
-      cuda_allocator
-      cuda_managed_allocator
-      pinned_allocator
-      cuda_device_guard
-      thread_local_allocator
-      stream_safe_cuda_allocator
-      device_context)
-  if(CUDA_VERSION GREATER_EQUAL 10.2)
-    list(APPEND AllocatorFacadeDeps cuda_virtual_mem_allocator)
-  endif()
-elseif(WITH_XPU)
-  set(AllocatorFacadeDeps xpu_info)
-elseif(WITH_IPU)
-  set(AllocatorFacadeDeps ipu_info)
-elseif(WITH_ASCEND)
-  set(AllocatorFacadeDeps ascend_npu_info)
-else()
-  set(AllocatorFacadeDeps)
+if(WITH_GPU)
+  nv_test(
+    thread_local_allocator_test
+    SRCS thread_local_allocator_test.cc
+    DEPS allocator)
 endif()
-
-if(WITH_CUSTOM_DEVICE)
-  cc_library(
-    custom_allocator
-    SRCS custom_allocator.cc
-    DEPS allocator device_manager)
-  set(AllocatorFacadeDeps ${AllocatorFacadeDeps} custom_allocator)
+if(WITH_ROCM)
+  hip_test(
+    thread_local_allocator_test
+    SRCS thread_local_allocator_test.cc
+    DEPS allocator)
 endif()
 
 if(WITH_GPU)
   nv_test(
     best_fit_allocator_test
     SRCS best_fit_allocator_test.cc best_fit_allocator_test.cu
-    DEPS best_fit_allocator locked_allocator cpu_allocator cuda_allocator
-         device_context memcpy)
+    DEPS allocator memcpy)
 elseif(WITH_ROCM)
   hip_test(
     best_fit_allocator_test
     SRCS best_fit_allocator_test.cc best_fit_allocator_test.cu
-    DEPS best_fit_allocator locked_allocator cpu_allocator cuda_allocator
-         device_context memcpy)
+    DEPS allocator memcpy)
 else()
   cc_test(
     best_fit_allocator_test
     SRCS best_fit_allocator_test.cc
-    DEPS best_fit_allocator locked_allocator cpu_allocator)
-endif()
-
-list(
-  APPEND
-  AllocatorFacadeDeps
-  cpu_allocator
-  locked_allocator
-  aligned_allocator
-  retry_allocator
-  buffered_allocator
-  naive_best_fit_allocator
-  auto_growth_best_fit_allocator
-  virtual_memory_auto_growth_best_fit_allocator
-  best_fit_allocator)
-
-if(WITH_ASCEND_CL)
-  list(APPEND AllocatorFacadeDeps npu_pinned_allocator)
+    DEPS allocator)
 endif()
 
-cc_library(
-  aligned_allocator
-  SRCS aligned_allocator.cc
-  DEPS allocator)
 cc_test(
   test_aligned_allocator
   SRCS test_aligned_allocator.cc
-  DEPS aligned_allocator)
-cc_library(
-  allocator_strategy
-  SRCS allocator_strategy.cc
-  DEPS gflags ${AllocatorFacadeDeps})
-cc_library(
-  allocator_facade
-  SRCS allocator_facade.cc
-  DEPS allocator_strategy stats)
-
-if(WITH_GPU)
-  target_link_libraries(allocator_facade cuda_graph)
-endif()
+  DEPS allocator)
 
 cc_test(
   retry_allocator_test
   SRCS retry_allocator_test.cc
-  DEPS retry_allocator locked_allocator cpu_allocator)
-if(WITH_TESTING)
-  if((WITH_GPU OR WITH_ROCM) AND TARGET retry_allocator_test)
-    target_link_libraries(retry_allocator_test cuda_allocator)
-  endif()
-
-  if(TEST retry_allocator_test)
+  DEPS allocator)
+if(TEST retry_allocator_test)
   set_tests_properties(retry_allocator_test PROPERTIES LABELS
                                                        "RUN_TYPE=EXCLUSIVE")
-  endif()
 endif()
 
 cc_test(
   allocator_facade_abs_flags_test
   SRCS allocator_facade_abs_flags_test.cc
-  DEPS allocator_facade)
+  DEPS allocator)
 
 cc_test(
   allocator_facade_frac_flags_test
   SRCS allocator_facade_frac_flags_test.cc
-  DEPS allocator_facade)
+  DEPS allocator)
 
-cc_library(
-  auto_growth_best_fit_allocator
-  SRCS auto_growth_best_fit_allocator.cc
-  DEPS allocator aligned_allocator flags)
 cc_test(
   auto_growth_best_fit_allocator_facade_test
   SRCS auto_growth_best_fit_allocator_facade_test.cc
-  DEPS cpu_allocator auto_growth_best_fit_allocator)
+  DEPS allocator)
 cc_test(
   auto_growth_best_fit_allocator_test
   SRCS auto_growth_best_fit_allocator_test.cc
-  DEPS auto_growth_best_fit_allocator)
-
-cc_library(
-  virtual_memory_auto_growth_best_fit_allocator
-  SRCS virtual_memory_auto_growth_best_fit_allocator.cc
-  DEPS allocator aligned_allocator)
+  DEPS allocator)
 
 if(NOT WIN32)
-  cc_library(
-    mmap_allocator
-    SRCS mmap_allocator.cc
-    DEPS allocator)
   cc_test(
     mmap_allocator_test
     SRCS mmap_allocator_test.cc
-    DEPS mmap_allocator allocator)
-  if(WITH_GPU)
-    cc_library(
-      cuda_ipc_allocator
-      SRCS cuda_ipc_allocator.cc
     DEPS allocator)
+endif()
+
+cc_test(
+  system_allocator_test
+  SRCS system_allocator_test.cc
+  DEPS allocator)
+
+cc_test(
+  buddy_allocator_test
+  SRCS buddy_allocator_test.cc
+  DEPS allocator)
+
+if(WITH_TESTING)
+  if(TEST buddy_allocator_test)
+    set_tests_properties(buddy_allocator_test PROPERTIES LABELS
+                                                         "RUN_TYPE=EXCLUSIVE")
+  endif()
+
+  # TODO(zhiqiu): why not win32? because wget is not found on windows
+  if(NOT WIN32)
+    add_custom_target(
+      download_data
+      COMMAND wget -nc
+              https://paddle-ci.cdn.bcebos.com/buddy_allocator_test_data.tar
+      COMMAND tar -xf buddy_allocator_test_data.tar)
+    add_dependencies(buddy_allocator_test download_data)
   endif()
 endif()

paddle/fluid/memory/detail/buddy_allocator.cc → paddle/fluid/memory/allocation/buddy_allocator.cc

@@ -12,7 +12,7 @@ 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/memory/detail/buddy_allocator.h"
+#include "paddle/fluid/memory/allocation/buddy_allocator.h"
 
 #include <algorithm>
 

paddle/fluid/memory/detail/buddy_allocator.h → paddle/fluid/memory/allocation/buddy_allocator.h

@@ -25,8 +25,8 @@ limitations under the License. */
 #include <utility>
 #include <vector>
 
-#include "paddle/fluid/memory/detail/memory_block.h"
-#include "paddle/fluid/memory/detail/system_allocator.h"
+#include "paddle/fluid/memory/allocation/memory_block.h"
+#include "paddle/fluid/memory/allocation/system_allocator.h"
 #include "paddle/fluid/platform/cpu_info.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/device/npu/npu_info.h"

paddle/fluid/memory/detail/buddy_allocator_test.cc → paddle/fluid/memory/allocation/buddy_allocator_test.cc

@@ -12,7 +12,7 @@ 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/memory/detail/buddy_allocator.h"
+#include "paddle/fluid/memory/allocation/buddy_allocator.h"
 
 #include <memory>
 
@@ -330,7 +330,7 @@ TEST(BuddyAllocator, SpeedAna) {
   std::vector<bool> vec_free_flag;
 
   std::string line;
-  int size, id;
+  int size = 0, id = 0;
   while (in_file >> size >> id) {
     vec_size.push_back(size);
     vec_pos.push_back(id);

paddle/fluid/memory/detail/memory_block.cc → paddle/fluid/memory/allocation/memory_block.cc

@@ -12,7 +12,7 @@ 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/memory/detail/memory_block.h"
+#include "paddle/fluid/memory/allocation/memory_block.h"
 
 #include "paddle/fluid/platform/enforce.h"
 

paddle/fluid/memory/detail/memory_block_desc.cc → paddle/fluid/memory/allocation/memory_block_desc.cc

@@ -15,7 +15,7 @@ limitations under the License. */
 #include <cstddef>
 #include <functional>
 
-#include "paddle/fluid/memory/detail/memory_block.h"
+#include "paddle/fluid/memory/allocation/memory_block.h"
 
 namespace paddle {
 namespace memory {

paddle/fluid/memory/detail/meta_cache.cc → paddle/fluid/memory/allocation/meta_cache.cc

@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "glog/logging.h"
-#include "paddle/fluid/memory/detail/memory_block.h"
+#include "paddle/fluid/memory/allocation/memory_block.h"
 #include "paddle/fluid/platform/enforce.h"
 
 namespace paddle {

paddle/fluid/memory/allocation/naive_best_fit_allocator.cc

@@ -18,8 +18,8 @@
 
 #include "gflags/gflags.h"
 #include "glog/logging.h"
-#include "paddle/fluid/memory/detail/buddy_allocator.h"
-#include "paddle/fluid/memory/detail/system_allocator.h"
+#include "paddle/fluid/memory/allocation/buddy_allocator.h"
+#include "paddle/fluid/memory/allocation/system_allocator.h"
 #include "paddle/fluid/platform/device/device_wrapper.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/enforce.h"

paddle/fluid/memory/detail/system_allocator.cc → paddle/fluid/memory/allocation/system_allocator.cc

@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 #define GLOG_NO_ABBREVIATED_SEVERITIES
 
-#include "paddle/fluid/memory/detail/system_allocator.h"
+#include "paddle/fluid/memory/allocation/system_allocator.h"
 
 #include "paddle/fluid/memory/stats.h"
 

paddle/fluid/memory/detail/system_allocator_test.cc → paddle/fluid/memory/allocation/system_allocator_test.cc

@@ -12,7 +12,7 @@ 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/memory/detail/system_allocator.h"
+#include "paddle/fluid/memory/allocation/system_allocator.h"
 
 #include <memory>
 

paddle/fluid/memory/allocation/thread_local_allocator.h

@@ -18,8 +18,8 @@
 #include <vector>
 
 #include "paddle/fluid/memory/allocation/allocator.h"
-#include "paddle/fluid/memory/detail/buddy_allocator.h"
-#include "paddle/fluid/memory/detail/system_allocator.h"
+#include "paddle/fluid/memory/allocation/buddy_allocator.h"
+#include "paddle/fluid/memory/allocation/system_allocator.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 
 namespace paddle {

paddle/fluid/memory/detail/CMakeLists.txt

-include(ExternalProject)
-
-cc_library(
-  memory_block
-  SRCS memory_block.cc memory_block_desc.cc meta_cache.cc
-  DEPS place)
-
-if(WITH_GPU)
-  nv_library(
-    system_allocator
-    SRCS system_allocator.cc
-    DEPS gflags cpu_info gpu_info place)
-elseif(WITH_ROCM)
-  hip_library(
-    system_allocator
-    SRCS system_allocator.cc
-    DEPS gflags cpu_info gpu_info place)
-elseif(${WITH_ASCEND_CL})
-  cc_library(
-    system_allocator
-    SRCS system_allocator.cc
-    DEPS gflags cpu_info npu_info place)
-elseif(WITH_MLU)
-  cc_library(
-    system_allocator
-    SRCS system_allocator.cc
-    DEPS gflags cpu_info mlu_info place)
-else()
-  cc_library(
-    system_allocator
-    SRCS system_allocator.cc
-    DEPS gflags cpu_info place)
-endif()
-
-cc_test(
-  system_allocator_test
-  SRCS system_allocator_test.cc
-  DEPS system_allocator)
-
-cc_library(
-  buddy_allocator
-  SRCS buddy_allocator.cc
-  DEPS memory_block system_allocator glog)
-
-cc_test(
-  buddy_allocator_test
-  SRCS buddy_allocator_test.cc
-  DEPS buddy_allocator)
-
-function(file_download_and_uncompress URL NAME)
-  message(STATUS "Download dependence[${NAME}] from ${URL}")
-  set(${NAME}_INCLUDE_DIR
-      ${THIRD_PARTY_PATH}/${NAME}
-      PARENT_SCOPE)
-  ExternalProject_Add(
-    extern_download_${NAME}
-    ${EXTERNAL_PROJECT_LOG_ARGS}
-    PREFIX ${THIRD_PARTY_PATH}/${NAME}
-    URL ${URL}
-    DOWNLOAD_DIR ${THIRD_PARTY_PATH}/${NAME}
-    SOURCE_DIR ${THIRD_PARTY_PATH}/${NAME}
-    DOWNLOAD_NO_PROGRESS 1
-    CONFIGURE_COMMAND ""
-    BUILD_COMMAND ""
-    UPDATE_COMMAND ""
-    INSTALL_COMMAND "")
-  set(third_party_deps
-      ${third_party_deps} extern_download_${NAME}
-      PARENT_SCOPE)
-endfunction()
-
-if(WITH_TESTING)
-  if(TEST buddy_allocator_test)
-    set_tests_properties(buddy_allocator_test PROPERTIES LABELS
-                                                         "RUN_TYPE=EXCLUSIVE")
-  endif()
-  set(URL "https://paddle-ci.cdn.bcebos.com/buddy_allocator_test_data.tar")
-  file_download_and_uncompress(URL "buddy_allocator")
-endif()

paddle/fluid/memory/pinned_memory_test.cu

@@ -15,7 +15,7 @@ limitations under the License. */
 
 #include <unordered_map>
 
-#include "paddle/fluid/memory/detail/memory_block.h"
+#include "paddle/fluid/memory/allocation/memory_block.h"
 #include "paddle/fluid/memory/memcpy.h"
 #include "paddle/fluid/memory/memory.h"
 #include "paddle/fluid/platform/cpu_info.h"

paddle/fluid/platform/CMakeLists.txt

@@ -88,12 +88,12 @@ if(WITH_GPU)
   nv_library(
     cuda_graph_with_memory_pool
     SRCS cuda_graph_with_memory_pool.cc
-    DEPS device_context allocator_facade cuda_graph)
+    DEPS device_context allocator cuda_graph)
 else()
   cc_library(
     cuda_graph_with_memory_pool
     SRCS cuda_graph_with_memory_pool.cc
-    DEPS device_context allocator_facade)
+    DEPS device_context allocator)
 endif()
 
 cc_library(

paddle/fluid/platform/device/gpu/cuda/CMakeLists.txt

 nv_library(
   cuda_graph
   SRCS cuda_graph.cc
-  DEPS enforce allocator_facade)
+  DEPS enforce)
 nv_library(
   cuda_profiler
   SRCS cuda_profiler.cc

paddle/fluid/platform/profiler.cc

@@ -310,8 +310,10 @@ RecordOpInfoSupplement::RecordOpInfoSupplement(
 
 std::map<const char *, std::map<uint64_t, std::vector<uint64_t>>>
     RecordMemEvent::size_cache;
+
 std::map<const char *, std::map<uint64_t, bool>>
     RecordMemEvent::has_initialized;
+
 RecordMemEvent::RecordMemEvent(const void *ptr,
                                const phi::Place &place,
                                size_t size,

paddle/fluid/platform/profiler.h

@@ -28,7 +28,6 @@ limitations under the License. */
 #include "paddle/fluid/platform/enforce.h"
 #include "paddle/fluid/platform/event.h"
 #include "paddle/fluid/platform/place.h"
-#include "paddle/fluid/platform/profiler.pb.h"
 #include "paddle/fluid/platform/profiler/event_tracing.h"
 #include "paddle/fluid/platform/profiler/mem_tracing.h"
 #include "paddle/fluid/platform/profiler/supplement_tracing.h"
@@ -39,6 +38,10 @@ limitations under the License. */
 namespace paddle {
 namespace platform {
 
+namespace proto {
+class Profile;
+}
+
 const int kEnableProfiler = 1;
 const int kDisableProfiler = 2;
 

paddle/fluid/pybind/CMakeLists.txt

@@ -84,10 +84,6 @@ endif()
 
 if(NOT WIN32)
   set(PYBIND_DEPS ${PYBIND_DEPS} data_loader)
-  set(PYBIND_DEPS ${PYBIND_DEPS} mmap_allocator)
-  if(WITH_GPU)
-    set(PYBIND_DEPS ${PYBIND_DEPS} cuda_ipc_allocator)
-  endif()
   if(WITH_NCCL OR WITH_RCCL)
     set(PYBIND_DEPS ${PYBIND_DEPS} nccl_context)
     set(PYBIND_DEPS ${PYBIND_DEPS} heter_ccl_context)