Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
大约 1 年 前同步成功

通知 695
Star 11112
Fork 2696
P
PaddleDetection
提交 be04d99f 编写于 作者: M minqiyang
浏览文件
操作

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into revert_vlog 

test=develop
上级 53433d7f 05b7ee7e
隐藏空白更改
内联 并排
Showing with 157 addition and 139 deletion
cmake/cuda.cmake
浏览文件 @ be04d99f
...@@ -199,8 +199,11 @@ elseif(CMAKE_BUILD_TYPE STREQUAL "MinSizeRel") ...@@ -199,8 +199,11 @@ elseif(CMAKE_BUILD_TYPE STREQUAL "MinSizeRel")
list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELEASE}) list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELEASE})
endif() endif()
else(NOT WIN32) else(NOT WIN32)
list(APPEND CUDA_NVCC_FLAGS "--compiler-options;/bigobj")
if(CMAKE_BUILD_TYPE STREQUAL "Debug") if(CMAKE_BUILD_TYPE STREQUAL "Debug")
list(APPEND CUDA_NVCC_FLAGS "-g -G") list(APPEND CUDA_NVCC_FLAGS "-g -G")
# match the cl's _ITERATOR_DEBUG_LEVEL
list(APPEND CUDA_NVCC_FLAGS "-D_DEBUG")
elseif(CMAKE_BUILD_TYPE STREQUAL "Release") elseif(CMAKE_BUILD_TYPE STREQUAL "Release")
list(APPEND CUDA_NVCC_FLAGS "-O3 -DNDEBUG") list(APPEND CUDA_NVCC_FLAGS "-O3 -DNDEBUG")
else() else()
......
cmake/external/pybind11.cmake
浏览文件 @ be04d99f
...@@ -26,7 +26,7 @@ ExternalProject_Add( ...@@ -26,7 +26,7 @@ ExternalProject_Add(
extern_pybind extern_pybind
${EXTERNAL_PROJECT_LOG_ARGS} ${EXTERNAL_PROJECT_LOG_ARGS}
GIT_REPOSITORY "https://github.com/pybind/pybind11.git" GIT_REPOSITORY "https://github.com/pybind/pybind11.git"
GIT_TAG "v2.1.1" GIT_TAG "v2.2.4"
PREFIX ${PYBIND_SOURCE_DIR} PREFIX ${PYBIND_SOURCE_DIR}
UPDATE_COMMAND "" UPDATE_COMMAND ""
CONFIGURE_COMMAND "" CONFIGURE_COMMAND ""
......
cmake/generic.cmake
浏览文件 @ be04d99f
...@@ -349,10 +349,17 @@ function(cc_test TARGET_NAME) ...@@ -349,10 +349,17 @@ function(cc_test TARGET_NAME)
set(oneValueArgs "") set(oneValueArgs "")
set(multiValueArgs SRCS DEPS ARGS) set(multiValueArgs SRCS DEPS ARGS)
cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN}) cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
if(WIN32)
list(APPEND win32_deps shlwapi)
if("${cc_test_DEPS};" MATCHES "python;")
list(REMOVE_ITEM cc_test_DEPS python)
list(APPEND win32_deps ${PYTHON_LIBRARIES})
endif()
endif(WIN32)
add_executable(${TARGET_NAME} ${cc_test_SRCS}) add_executable(${TARGET_NAME} ${cc_test_SRCS})
target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog) target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
if(WIN32) if(WIN32)
target_link_libraries(${TARGET_NAME} shlwapi) target_link_libraries(${TARGET_NAME} ${win32_deps})
endif(WIN32) endif(WIN32)
add_dependencies(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog) add_dependencies(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
add_test(NAME ${TARGET_NAME} add_test(NAME ${TARGET_NAME}
...@@ -683,7 +690,7 @@ function(py_test TARGET_NAME) ...@@ -683,7 +690,7 @@ function(py_test TARGET_NAME)
set(multiValueArgs SRCS DEPS ARGS ENVS) set(multiValueArgs SRCS DEPS ARGS ENVS)
cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN}) cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
add_test(NAME ${TARGET_NAME} add_test(NAME ${TARGET_NAME}
COMMAND env FLAGS_init_allocated_mem=true FLAGS_cudnn_deterministic=true COMMAND ${CMAKE_COMMAND} -E env FLAGS_init_allocated_mem=true FLAGS_cudnn_deterministic=true
FLAGS_cpu_deterministic=true FLAGS_cpu_deterministic=true
PYTHONPATH=${PADDLE_BINARY_DIR}/python ${py_test_ENVS} PYTHONPATH=${PADDLE_BINARY_DIR}/python ${py_test_ENVS}
${PYTHON_EXECUTABLE} -u ${py_test_SRCS} ${py_test_ARGS} ${PYTHON_EXECUTABLE} -u ${py_test_SRCS} ${py_test_ARGS}
......
paddle/fluid/API.spec
浏览文件 @ be04d99f
...@@ -26,10 +26,10 @@ paddle.fluid.release_memory ArgSpec(args=['input_program', 'skip_opt_set'], vara ...@@ -26,10 +26,10 @@ paddle.fluid.release_memory ArgSpec(args=['input_program', 'skip_opt_set'], vara
paddle.fluid.DistributeTranspilerConfig.__init__ paddle.fluid.DistributeTranspilerConfig.__init__
paddle.fluid.ParallelExecutor.__init__ ArgSpec(args=['self', 'use_cuda', 'loss_name', 'main_program', 'share_vars_from', 'exec_strategy', 'build_strategy', 'num_trainers', 'trainer_id', 'scope'], varargs=None, keywords=None, defaults=(None, None, None, None, None, 1, 0, None)) paddle.fluid.ParallelExecutor.__init__ ArgSpec(args=['self', 'use_cuda', 'loss_name', 'main_program', 'share_vars_from', 'exec_strategy', 'build_strategy', 'num_trainers', 'trainer_id', 'scope'], varargs=None, keywords=None, defaults=(None, None, None, None, None, 1, 0, None))
paddle.fluid.ParallelExecutor.run ArgSpec(args=['self', 'fetch_list', 'feed', 'feed_dict', 'return_numpy'], varargs=None, keywords=None, defaults=(None, None, True)) paddle.fluid.ParallelExecutor.run ArgSpec(args=['self', 'fetch_list', 'feed', 'feed_dict', 'return_numpy'], varargs=None, keywords=None, defaults=(None, None, True))
paddle.fluid.ExecutionStrategy.__init__ __init__(self: paddle.fluid.core.ExecutionStrategy) -> None paddle.fluid.ExecutionStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.ExecutionStrategy) -> None
paddle.fluid.BuildStrategy.GradientScaleStrategy.__init__ __init__(self: paddle.fluid.core.GradientScaleStrategy, arg0: int) -> None paddle.fluid.BuildStrategy.GradientScaleStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.BuildStrategy.GradientScaleStrategy, arg0: int) -> None
paddle.fluid.BuildStrategy.ReduceStrategy.__init__ __init__(self: paddle.fluid.core.ReduceStrategy, arg0: int) -> None paddle.fluid.BuildStrategy.ReduceStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.BuildStrategy.ReduceStrategy, arg0: int) -> None
paddle.fluid.BuildStrategy.__init__ __init__(self: paddle.fluid.core.BuildStrategy) -> None paddle.fluid.BuildStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.BuildStrategy) -> None
paddle.fluid.create_lod_tensor ArgSpec(args=['data', 'recursive_seq_lens', 'place'], varargs=None, keywords=None, defaults=None) paddle.fluid.create_lod_tensor ArgSpec(args=['data', 'recursive_seq_lens', 'place'], varargs=None, keywords=None, defaults=None)
paddle.fluid.create_random_int_lodtensor ArgSpec(args=['recursive_seq_lens', 'base_shape', 'place', 'low', 'high'], varargs=None, keywords=None, defaults=None) paddle.fluid.create_random_int_lodtensor ArgSpec(args=['recursive_seq_lens', 'base_shape', 'place', 'low', 'high'], varargs=None, keywords=None, defaults=None)
paddle.fluid.io.save_vars ArgSpec(args=['executor', 'dirname', 'main_program', 'vars', 'predicate', 'filename'], varargs=None, keywords=None, defaults=(None, None, None, None)) paddle.fluid.io.save_vars ArgSpec(args=['executor', 'dirname', 'main_program', 'vars', 'predicate', 'filename'], varargs=None, keywords=None, defaults=(None, None, None, None))
......
paddle/fluid/framework/CMakeLists.txt
浏览文件 @ be04d99f
...@@ -116,14 +116,9 @@ cc_test(op_proto_maker_test SRCS op_proto_maker_test.cc DEPS op_proto_maker) ...@@ -116,14 +116,9 @@ cc_test(op_proto_maker_test SRCS op_proto_maker_test.cc DEPS op_proto_maker)
cc_library(op_info SRCS op_info.cc DEPS attribute framework_proto) cc_library(op_info SRCS op_info.cc DEPS attribute framework_proto)
cc_library(shape_inference SRCS shape_inference.cc DEPS ddim attribute device_context) cc_library(shape_inference SRCS shape_inference.cc DEPS ddim attribute device_context)
if (NOT WIN32)
cc_library(transfer_scope_cache SRCS transfer_scope_cache.cc DEPS scope framework_proto device_context) cc_library(transfer_scope_cache SRCS transfer_scope_cache.cc DEPS scope framework_proto device_context)
cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog
shape_inference data_transform lod_tensor profiler transfer_scope_cache) shape_inference data_transform lod_tensor profiler transfer_scope_cache)
else()
cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog
shape_inference data_transform lod_tensor)
endif(NOT WIN32)
cc_test(operator_test SRCS operator_test.cc DEPS operator op_registry device_context) cc_test(operator_test SRCS operator_test.cc DEPS operator op_registry device_context)
......
paddle/fluid/framework/details/all_reduce_op_handle.cc
浏览文件 @ be04d99f
...@@ -23,7 +23,7 @@ namespace paddle { ...@@ -23,7 +23,7 @@ namespace paddle {
namespace framework { namespace framework {
namespace details { namespace details {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
AllReduceOpHandle::AllReduceOpHandle(ir::Node *node, AllReduceOpHandle::AllReduceOpHandle(ir::Node *node,
const std::vector<Scope *> &local_scopes, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
...@@ -74,7 +74,7 @@ void AllReduceOpHandle::RunImpl() { ...@@ -74,7 +74,7 @@ void AllReduceOpHandle::RunImpl() {
} }
if (platform::is_gpu_place(lod_tensors[0]->place())) { if (platform::is_gpu_place(lod_tensors[0]->place())) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
PADDLE_ENFORCE(nccl_ctxs_, "nccl_ctxs should not be nullptr."); PADDLE_ENFORCE(nccl_ctxs_, "nccl_ctxs should not be nullptr.");
int dtype = -1; int dtype = -1;
size_t numel = 0; size_t numel = 0;
......
paddle/fluid/framework/details/all_reduce_op_handle.h
浏览文件 @ be04d99f
...@@ -20,7 +20,7 @@ ...@@ -20,7 +20,7 @@
#include "paddle/fluid/framework/details/op_handle_base.h" #include "paddle/fluid/framework/details/op_handle_base.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/scope.h" #include "paddle/fluid/framework/scope.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -29,7 +29,7 @@ namespace framework { ...@@ -29,7 +29,7 @@ namespace framework {
namespace details { namespace details {
struct AllReduceOpHandle : public OpHandleBase { struct AllReduceOpHandle : public OpHandleBase {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
AllReduceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes, AllReduceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
const platform::NCCLContextMap *ctxs); const platform::NCCLContextMap *ctxs);
...@@ -49,7 +49,7 @@ struct AllReduceOpHandle : public OpHandleBase { ...@@ -49,7 +49,7 @@ struct AllReduceOpHandle : public OpHandleBase {
private: private:
std::vector<Scope *> local_scopes_; std::vector<Scope *> local_scopes_;
std::vector<platform::Place> places_; std::vector<platform::Place> places_;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
const platform::NCCLContextMap *nccl_ctxs_; const platform::NCCLContextMap *nccl_ctxs_;
#endif #endif
}; };
......
paddle/fluid/framework/details/broadcast_op_handle.cc
浏览文件 @ be04d99f
...@@ -82,7 +82,7 @@ void BroadcastOpHandle::BroadcastOneVar( ...@@ -82,7 +82,7 @@ void BroadcastOpHandle::BroadcastOneVar(
}); });
} }
} else { } else {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
VarHandle *out_handle = nullptr; VarHandle *out_handle = nullptr;
int root_id = boost::get<platform::CUDAPlace>(in_tensor.place()).device; int root_id = boost::get<platform::CUDAPlace>(in_tensor.place()).device;
std::vector<std::function<void()>> broadcast_calls; std::vector<std::function<void()>> broadcast_calls;
......
paddle/fluid/framework/details/broadcast_op_handle.h
浏览文件 @ be04d99f
...@@ -24,7 +24,7 @@ ...@@ -24,7 +24,7 @@
#include "paddle/fluid/framework/selected_rows.h" #include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -34,7 +34,7 @@ namespace details { ...@@ -34,7 +34,7 @@ namespace details {
struct BroadcastOpHandle : public OpHandleBase { struct BroadcastOpHandle : public OpHandleBase {
public: public:
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
BroadcastOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes, BroadcastOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
const platform::NCCLContextMap *nccl_ctxs) const platform::NCCLContextMap *nccl_ctxs)
...@@ -68,7 +68,7 @@ struct BroadcastOpHandle : public OpHandleBase { ...@@ -68,7 +68,7 @@ struct BroadcastOpHandle : public OpHandleBase {
std::vector<Scope *> local_scopes_; std::vector<Scope *> local_scopes_;
std::vector<platform::Place> places_; std::vector<platform::Place> places_;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
const platform::NCCLContextMap *nccl_ctxs_; const platform::NCCLContextMap *nccl_ctxs_;
#endif #endif
......
paddle/fluid/framework/details/broadcast_op_handle_test.h
浏览文件 @ be04d99f
...@@ -42,7 +42,7 @@ struct TestBroadcastOpHandle { ...@@ -42,7 +42,7 @@ struct TestBroadcastOpHandle {
std::vector<std::unique_ptr<ir::Node>> nodes_; std::vector<std::unique_ptr<ir::Node>> nodes_;
std::vector<p::Place> place_list_; std::vector<p::Place> place_list_;
bool use_gpu_; bool use_gpu_;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
std::unique_ptr<platform::NCCLContextMap> nccl_ctxs_; std::unique_ptr<platform::NCCLContextMap> nccl_ctxs_;
#endif #endif
...@@ -50,7 +50,7 @@ struct TestBroadcastOpHandle { ...@@ -50,7 +50,7 @@ struct TestBroadcastOpHandle {
for (size_t j = 0; j < ctxs_.size(); ++j) { for (size_t j = 0; j < ctxs_.size(); ++j) {
ctxs_[j]->Wait(); ctxs_[j]->Wait();
} }
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
if (nccl_ctxs_) { if (nccl_ctxs_) {
nccl_ctxs_->WaitAll(); nccl_ctxs_->WaitAll();
} }
...@@ -60,7 +60,7 @@ struct TestBroadcastOpHandle { ...@@ -60,7 +60,7 @@ struct TestBroadcastOpHandle {
void InitCtxOnGpu(bool use_gpu) { void InitCtxOnGpu(bool use_gpu) {
use_gpu_ = use_gpu; use_gpu_ = use_gpu;
if (use_gpu_) { if (use_gpu_) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
int count = p::GetCUDADeviceCount(); int count = p::GetCUDADeviceCount();
if (count <= 1) { if (count <= 1) {
LOG(WARNING) << "Cannot test multi-gpu Broadcast, because the CUDA " LOG(WARNING) << "Cannot test multi-gpu Broadcast, because the CUDA "
...@@ -84,7 +84,7 @@ struct TestBroadcastOpHandle { ...@@ -84,7 +84,7 @@ struct TestBroadcastOpHandle {
place_list_.push_back(p); place_list_.push_back(p);
ctxs_.emplace_back(new p::CPUDeviceContext(p)); ctxs_.emplace_back(new p::CPUDeviceContext(p));
} }
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
nccl_ctxs_.reset(nullptr); nccl_ctxs_.reset(nullptr);
#endif #endif
} }
...@@ -106,14 +106,14 @@ struct TestBroadcastOpHandle { ...@@ -106,14 +106,14 @@ struct TestBroadcastOpHandle {
nodes_.emplace_back( nodes_.emplace_back(
ir::CreateNodeForTest("node0", ir::Node::Type::kOperation)); ir::CreateNodeForTest("node0", ir::Node::Type::kOperation));
if (use_gpu_) { if (use_gpu_) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_ = new BroadcastOpHandle(nodes_.back().get(), local_scopes_, op_handle_ = new BroadcastOpHandle(nodes_.back().get(), local_scopes_,
place_list_, nccl_ctxs_.get()); place_list_, nccl_ctxs_.get());
#else #else
PADDLE_THROW("CUDA is not support."); PADDLE_THROW("CUDA is not support.");
#endif #endif
} else { } else {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_ = new BroadcastOpHandle(nodes_.back().get(), local_scopes_, op_handle_ = new BroadcastOpHandle(nodes_.back().get(), local_scopes_,
place_list_, nccl_ctxs_.get()); place_list_, nccl_ctxs_.get());
#else #else
......
paddle/fluid/framework/details/build_strategy.cc
浏览文件 @ be04d99f
...@@ -96,7 +96,7 @@ std::unique_ptr<ir::Graph> BuildStrategy::Apply( ...@@ -96,7 +96,7 @@ std::unique_ptr<ir::Graph> BuildStrategy::Apply(
const std::string &loss_var_name, const std::string &loss_var_name,
const std::unordered_set<std::string> &param_names, const std::unordered_set<std::string> &param_names,
const std::vector<Scope *> &local_scopes, const std::vector<Scope *> &local_scopes,
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
const bool use_cuda, platform::NCCLContextMap *nccl_ctxs) const { const bool use_cuda, platform::NCCLContextMap *nccl_ctxs) const {
#else #else
const bool use_cuda) const { const bool use_cuda) const {
...@@ -118,7 +118,7 @@ std::unique_ptr<ir::Graph> BuildStrategy::Apply( ...@@ -118,7 +118,7 @@ std::unique_ptr<ir::Graph> BuildStrategy::Apply(
pass->Erase("local_scopes"); pass->Erase("local_scopes");
pass->SetNotOwned<const std::vector<Scope *>>("local_scopes", pass->SetNotOwned<const std::vector<Scope *>>("local_scopes",
&local_scopes); &local_scopes);
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
platform::NCCLContextMap *nctx = use_cuda ? nccl_ctxs : nullptr; platform::NCCLContextMap *nctx = use_cuda ? nccl_ctxs : nullptr;
pass->Erase("nccl_ctxs"); pass->Erase("nccl_ctxs");
pass->SetNotOwned<platform::NCCLContextMap>("nccl_ctxs", nctx); pass->SetNotOwned<platform::NCCLContextMap>("nccl_ctxs", nctx);
......
paddle/fluid/framework/details/build_strategy.h
浏览文件 @ be04d99f
...@@ -23,7 +23,7 @@ ...@@ -23,7 +23,7 @@
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/enforce.h" #include "paddle/fluid/platform/enforce.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -98,7 +98,7 @@ struct BuildStrategy { ...@@ -98,7 +98,7 @@ struct BuildStrategy {
const std::string &loss_var_name, const std::string &loss_var_name,
const std::unordered_set<std::string> &param_names, const std::unordered_set<std::string> &param_names,
const std::vector<Scope *> &local_scopes, const std::vector<Scope *> &local_scopes,
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
const bool use_cuda, platform::NCCLContextMap *nccl_ctxs) const; const bool use_cuda, platform::NCCLContextMap *nccl_ctxs) const;
#else #else
const bool use_cuda) const; const bool use_cuda) const;
......
paddle/fluid/framework/details/data_balance_op_handle.cc
浏览文件 @ be04d99f
...@@ -20,7 +20,7 @@ namespace paddle { ...@@ -20,7 +20,7 @@ namespace paddle {
namespace framework { namespace framework {
namespace details { namespace details {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
DataBalanceOpHandle::DataBalanceOpHandle( DataBalanceOpHandle::DataBalanceOpHandle(
ir::Node *node, const std::vector<Scope *> &local_scopes, ir::Node *node, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
......
paddle/fluid/framework/details/data_balance_op_handle.h
浏览文件 @ be04d99f
...@@ -19,7 +19,7 @@ ...@@ -19,7 +19,7 @@
#include "paddle/fluid/framework/details/op_handle_base.h" #include "paddle/fluid/framework/details/op_handle_base.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/scope.h" #include "paddle/fluid/framework/scope.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -29,7 +29,7 @@ namespace details { ...@@ -29,7 +29,7 @@ namespace details {
struct DataBalanceOpHandle : public OpHandleBase { struct DataBalanceOpHandle : public OpHandleBase {
public: public:
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
DataBalanceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes, DataBalanceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
const platform::NCCLContextMap *ctxs); const platform::NCCLContextMap *ctxs);
......
paddle/fluid/framework/details/fused_broadcast_op_handle.h
浏览文件 @ be04d99f
...@@ -25,7 +25,7 @@ ...@@ -25,7 +25,7 @@
#include "paddle/fluid/framework/selected_rows.h" #include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -35,7 +35,7 @@ namespace details { ...@@ -35,7 +35,7 @@ namespace details {
struct FusedBroadcastOpHandle : public BroadcastOpHandle { struct FusedBroadcastOpHandle : public BroadcastOpHandle {
public: public:
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
FusedBroadcastOpHandle(ir::Node *node, FusedBroadcastOpHandle(ir::Node *node,
const std::vector<Scope *> local_scopes, const std::vector<Scope *> local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
......
paddle/fluid/framework/details/fused_broadcast_op_handle_test.cc
浏览文件 @ be04d99f
...@@ -44,14 +44,14 @@ struct TestFusedBroadcastOpHandle : TestBroadcastOpHandle { ...@@ -44,14 +44,14 @@ struct TestFusedBroadcastOpHandle : TestBroadcastOpHandle {
nodes_.emplace_back( nodes_.emplace_back(
ir::CreateNodeForTest("fused_broadcast", ir::Node::Type::kOperation)); ir::CreateNodeForTest("fused_broadcast", ir::Node::Type::kOperation));
if (use_gpu_) { if (use_gpu_) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_ = new FusedBroadcastOpHandle( op_handle_ = new FusedBroadcastOpHandle(
nodes_.back().get(), local_scopes_, place_list_, nccl_ctxs_.get()); nodes_.back().get(), local_scopes_, place_list_, nccl_ctxs_.get());
#else #else
PADDLE_THROW("CUDA is not supported."); PADDLE_THROW("CUDA is not supported.");
#endif #endif
} else { } else {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_ = new FusedBroadcastOpHandle( op_handle_ = new FusedBroadcastOpHandle(
nodes_.back().get(), local_scopes_, place_list_, nccl_ctxs_.get()); nodes_.back().get(), local_scopes_, place_list_, nccl_ctxs_.get());
#else #else
......
paddle/fluid/framework/details/multi_devices_graph_pass.cc
浏览文件 @ be04d99f
...@@ -142,7 +142,7 @@ void MultiDevSSAGraphBuilder::Init() const { ...@@ -142,7 +142,7 @@ void MultiDevSSAGraphBuilder::Init() const {
places_ = Get<const std::vector<platform::Place>>(kPlaces); places_ = Get<const std::vector<platform::Place>>(kPlaces);
local_scopes_ = Get<const std::vector<Scope *>>(kLocalScopes); local_scopes_ = Get<const std::vector<Scope *>>(kLocalScopes);
strategy_ = Get<const BuildStrategy>(kStrategy); strategy_ = Get<const BuildStrategy>(kStrategy);
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
nccl_ctxs_ = &Get<platform::NCCLContextMap>("nccl_ctxs"); nccl_ctxs_ = &Get<platform::NCCLContextMap>("nccl_ctxs");
#endif #endif
...@@ -431,7 +431,7 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilder::ApplyImpl( ...@@ -431,7 +431,7 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilder::ApplyImpl(
} }
} }
bool use_gpu = false; bool use_gpu = false;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
use_gpu = nccl_ctxs_ != nullptr; use_gpu = nccl_ctxs_ != nullptr;
#endif #endif
...@@ -478,7 +478,7 @@ bool MultiDevSSAGraphBuilder::IsSparseGradient(const std::string &og) const { ...@@ -478,7 +478,7 @@ bool MultiDevSSAGraphBuilder::IsSparseGradient(const std::string &og) const {
void MultiDevSSAGraphBuilder::SetCommunicationContext( void MultiDevSSAGraphBuilder::SetCommunicationContext(
OpHandleBase *op_handle, const platform::Place &p) const { OpHandleBase *op_handle, const platform::Place &p) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
if (nccl_ctxs_ == nullptr) { if (nccl_ctxs_ == nullptr) {
op_handle->SetDeviceContext(p, op_handle->SetDeviceContext(p,
platform::DeviceContextPool::Instance().Get(p)); platform::DeviceContextPool::Instance().Get(p));
...@@ -492,7 +492,7 @@ void MultiDevSSAGraphBuilder::SetCommunicationContext( ...@@ -492,7 +492,7 @@ void MultiDevSSAGraphBuilder::SetCommunicationContext(
void MultiDevSSAGraphBuilder::CreateBroadcastOp(ir::Graph *result, void MultiDevSSAGraphBuilder::CreateBroadcastOp(ir::Graph *result,
const std::string &p_name, const std::string &p_name,
size_t src_dev_id) const { size_t src_dev_id) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
auto *op_handle = new BroadcastOpHandle( auto *op_handle = new BroadcastOpHandle(
result->CreateEmptyNode("broadcast", ir::Node::Type::kOperation), result->CreateEmptyNode("broadcast", ir::Node::Type::kOperation),
local_scopes_, places_, nccl_ctxs_); local_scopes_, places_, nccl_ctxs_);
...@@ -522,7 +522,7 @@ void MultiDevSSAGraphBuilder::CreateBroadcastOp(ir::Graph *result, ...@@ -522,7 +522,7 @@ void MultiDevSSAGraphBuilder::CreateBroadcastOp(ir::Graph *result,
void MultiDevSSAGraphBuilder::CreateFusedBroadcastOp( void MultiDevSSAGraphBuilder::CreateFusedBroadcastOp(
ir::Graph *result, ir::Graph *result,
const std::vector<std::unordered_set<std::string>> &bcast_varnames) const { const std::vector<std::unordered_set<std::string>> &bcast_varnames) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
auto *op_handle = new FusedBroadcastOpHandle( auto *op_handle = new FusedBroadcastOpHandle(
result->CreateEmptyNode("fused_broadcast", ir::Node::Type::kOperation), result->CreateEmptyNode("fused_broadcast", ir::Node::Type::kOperation),
local_scopes_, places_, nccl_ctxs_); local_scopes_, places_, nccl_ctxs_);
...@@ -568,7 +568,7 @@ void MultiDevSSAGraphBuilder::CreateComputationalOp(ir::Graph *result, ...@@ -568,7 +568,7 @@ void MultiDevSSAGraphBuilder::CreateComputationalOp(ir::Graph *result,
void MultiDevSSAGraphBuilder::InsertAllReduceOp(ir::Graph *result, void MultiDevSSAGraphBuilder::InsertAllReduceOp(ir::Graph *result,
const std::string &og) const { const std::string &og) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
result->Get<GraphOps>(kGraphOps).emplace_back(new AllReduceOpHandle( result->Get<GraphOps>(kGraphOps).emplace_back(new AllReduceOpHandle(
result->CreateEmptyNode("allreduce", ir::Node::Type::kOperation), result->CreateEmptyNode("allreduce", ir::Node::Type::kOperation),
local_scopes_, places_, nccl_ctxs_)); local_scopes_, places_, nccl_ctxs_));
...@@ -597,7 +597,7 @@ void MultiDevSSAGraphBuilder::InsertAllReduceOp(ir::Graph *result, ...@@ -597,7 +597,7 @@ void MultiDevSSAGraphBuilder::InsertAllReduceOp(ir::Graph *result,
void MultiDevSSAGraphBuilder::InsertDataBalanceOp( void MultiDevSSAGraphBuilder::InsertDataBalanceOp(
ir::Graph *result, const std::vector<std::string> &datas) const { ir::Graph *result, const std::vector<std::string> &datas) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
result->Get<GraphOps>(kGraphOps).emplace_back(new DataBalanceOpHandle( result->Get<GraphOps>(kGraphOps).emplace_back(new DataBalanceOpHandle(
result->CreateEmptyNode("data_balance", ir::Node::Type::kOperation), result->CreateEmptyNode("data_balance", ir::Node::Type::kOperation),
local_scopes_, places_, nccl_ctxs_)); local_scopes_, places_, nccl_ctxs_));
...@@ -694,7 +694,7 @@ void MultiDevSSAGraphBuilder::CreateComputationalOps(ir::Graph *result, ...@@ -694,7 +694,7 @@ void MultiDevSSAGraphBuilder::CreateComputationalOps(ir::Graph *result,
VarHandle *MultiDevSSAGraphBuilder::CreateReduceOp(ir::Graph *result, VarHandle *MultiDevSSAGraphBuilder::CreateReduceOp(ir::Graph *result,
const std::string &og, const std::string &og,
int dst_dev_id) const { int dst_dev_id) const {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
result->Get<GraphOps>(kGraphOps).emplace_back(new ReduceOpHandle( result->Get<GraphOps>(kGraphOps).emplace_back(new ReduceOpHandle(
result->CreateEmptyNode("reduce", ir::Node::Type::kOperation), result->CreateEmptyNode("reduce", ir::Node::Type::kOperation),
local_scopes_, places_, nccl_ctxs_)); local_scopes_, places_, nccl_ctxs_));
......
paddle/fluid/framework/details/multi_devices_graph_pass.h
浏览文件 @ be04d99f
...@@ -40,7 +40,7 @@ class MultiDevSSAGraphBuilder : public ir::Pass { ...@@ -40,7 +40,7 @@ class MultiDevSSAGraphBuilder : public ir::Pass {
size_t device_id) const; size_t device_id) const;
void Init() const; void Init() const;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
mutable platform::NCCLContextMap *nccl_ctxs_; mutable platform::NCCLContextMap *nccl_ctxs_;
#endif #endif
......
paddle/fluid/framework/details/reduce_op_handle.cc
浏览文件 @ be04d99f
...@@ -125,7 +125,7 @@ void ReduceOpHandle::RunImpl() { ...@@ -125,7 +125,7 @@ void ReduceOpHandle::RunImpl() {
} }
}); });
} else if (paddle::platform::is_gpu_place(lod_tensors[0]->place())) { } else if (paddle::platform::is_gpu_place(lod_tensors[0]->place())) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
auto pre_in = pre_in_var->Get<framework::LoDTensor>(); auto pre_in = pre_in_var->Get<framework::LoDTensor>();
VariableVisitor::ShareDimsAndLoD(*pre_in_var, out_var); VariableVisitor::ShareDimsAndLoD(*pre_in_var, out_var);
VariableVisitor::GetMutableTensor(out_var).mutable_data( VariableVisitor::GetMutableTensor(out_var).mutable_data(
......
paddle/fluid/framework/details/reduce_op_handle.h
浏览文件 @ be04d99f
...@@ -23,7 +23,7 @@ ...@@ -23,7 +23,7 @@
#include "paddle/fluid/framework/scope.h" #include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/selected_rows.h" #include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
#endif #endif
...@@ -35,7 +35,7 @@ struct ReduceOpHandle : public OpHandleBase { ...@@ -35,7 +35,7 @@ struct ReduceOpHandle : public OpHandleBase {
std::vector<Scope *> local_scopes_; std::vector<Scope *> local_scopes_;
std::vector<platform::Place> places_; std::vector<platform::Place> places_;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
const platform::NCCLContextMap *nccl_ctxs_; const platform::NCCLContextMap *nccl_ctxs_;
ReduceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes, ReduceOpHandle(ir::Node *node, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
......
paddle/fluid/framework/details/reduce_op_handle_test.cc
浏览文件 @ be04d99f
...@@ -35,7 +35,7 @@ struct TestReduceOpHandle { ...@@ -35,7 +35,7 @@ struct TestReduceOpHandle {
std::vector<p::Place> gpu_list_; std::vector<p::Place> gpu_list_;
std::vector<std::unique_ptr<p::DeviceContext>> ctxs_; std::vector<std::unique_ptr<p::DeviceContext>> ctxs_;
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
std::unique_ptr<platform::NCCLContextMap> nccl_ctxs_; std::unique_ptr<platform::NCCLContextMap> nccl_ctxs_;
#endif #endif
...@@ -43,7 +43,7 @@ struct TestReduceOpHandle { ...@@ -43,7 +43,7 @@ struct TestReduceOpHandle {
for (size_t j = 0; j < ctxs_.size(); ++j) { for (size_t j = 0; j < ctxs_.size(); ++j) {
ctxs_[j]->Wait(); ctxs_[j]->Wait();
} }
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
if (nccl_ctxs_) { if (nccl_ctxs_) {
nccl_ctxs_->WaitAll(); nccl_ctxs_->WaitAll();
} }
...@@ -53,7 +53,7 @@ struct TestReduceOpHandle { ...@@ -53,7 +53,7 @@ struct TestReduceOpHandle {
void InitCtxOnGpu(bool use_gpu) { void InitCtxOnGpu(bool use_gpu) {
use_gpu_ = use_gpu; use_gpu_ = use_gpu;
if (use_gpu) { if (use_gpu) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
int count = p::GetCUDADeviceCount(); int count = p::GetCUDADeviceCount();
if (count <= 1) { if (count <= 1) {
LOG(WARNING) << "Cannot test multi-gpu Broadcast, because the CUDA " LOG(WARNING) << "Cannot test multi-gpu Broadcast, because the CUDA "
...@@ -77,7 +77,7 @@ struct TestReduceOpHandle { ...@@ -77,7 +77,7 @@ struct TestReduceOpHandle {
gpu_list_.push_back(p); gpu_list_.push_back(p);
ctxs_.emplace_back(new p::CPUDeviceContext(p)); ctxs_.emplace_back(new p::CPUDeviceContext(p));
} }
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
nccl_ctxs_.reset(nullptr); nccl_ctxs_.reset(nullptr);
#endif #endif
} }
...@@ -99,14 +99,14 @@ struct TestReduceOpHandle { ...@@ -99,14 +99,14 @@ struct TestReduceOpHandle {
nodes.emplace_back(new ir::Node("node")); nodes.emplace_back(new ir::Node("node"));
if (use_gpu_) { if (use_gpu_) {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_.reset(new ReduceOpHandle(nodes.back().get(), local_scopes_, op_handle_.reset(new ReduceOpHandle(nodes.back().get(), local_scopes_,
gpu_list_, nccl_ctxs_.get())); gpu_list_, nccl_ctxs_.get()));
#else #else
PADDLE_THROW("CUDA is not support."); PADDLE_THROW("CUDA is not support.");
#endif #endif
} else { } else {
#ifdef PADDLE_WITH_CUDA #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
op_handle_.reset(new ReduceOpHandle(nodes.back().get(), local_scopes_, op_handle_.reset(new ReduceOpHandle(nodes.back().get(), local_scopes_,
gpu_list_, nccl_ctxs_.get())); gpu_list_, nccl_ctxs_.get()));
#else #else
......
paddle/fluid/framework/ir/is_test_pass_tester.cc
浏览文件 @ be04d99f
...@@ -15,7 +15,10 @@ ...@@ -15,7 +15,10 @@
#include "paddle/fluid/framework/ir/is_test_pass.h" #include "paddle/fluid/framework/ir/is_test_pass.h"
#include <gtest/gtest.h> #include <gtest/gtest.h>
#ifdef _WIN32
#undef FALSE
#undef TRUE
#endif
namespace paddle { namespace paddle {
namespace framework { namespace framework {
namespace ir { namespace ir {
......
paddle/fluid/framework/lod_tensor.cc
浏览文件 @ be04d99f
...@@ -26,10 +26,8 @@ limitations under the License. */ ...@@ -26,10 +26,8 @@ limitations under the License. */
#include "paddle/fluid/memory/memcpy.h" #include "paddle/fluid/memory/memcpy.h"
#include "paddle/fluid/memory/memory.h" #include "paddle/fluid/memory/memory.h"
#if !defined(_WIN32)
#include "paddle/fluid/recordio/scanner.h" #include "paddle/fluid/recordio/scanner.h"
#include "paddle/fluid/recordio/writer.h" #include "paddle/fluid/recordio/writer.h"
#endif // _WIN32
namespace paddle { namespace paddle {
namespace framework { namespace framework {
...@@ -305,7 +303,6 @@ void DeserializeFromStream(std::istream &is, LoDTensor *tensor, ...@@ -305,7 +303,6 @@ void DeserializeFromStream(std::istream &is, LoDTensor *tensor,
TensorFromStream(is, static_cast<Tensor *>(tensor), dev_ctx); TensorFromStream(is, static_cast<Tensor *>(tensor), dev_ctx);
} }
#if !defined(_WIN32)
void WriteToRecordIO(recordio::Writer *writer, void WriteToRecordIO(recordio::Writer *writer,
const std::vector<LoDTensor> &tensor, const std::vector<LoDTensor> &tensor,
const platform::DeviceContext &dev_ctx) { const platform::DeviceContext &dev_ctx) {
...@@ -335,19 +332,7 @@ bool ReadFromRecordIO(recordio::Scanner *scanner, ...@@ -335,19 +332,7 @@ bool ReadFromRecordIO(recordio::Scanner *scanner,
return true; return true;
} }
#else
class Writer {};
class Scanner {};
void WriteToRecordIO(recordio::Writer *writer,
const std::vector<LoDTensor> &tensor,
const platform::DeviceContext &dev_ctx) {}
bool ReadFromRecordIO(recordio::Scanner *scanner,
const platform::DeviceContext &dev_ctx,
std::vector<LoDTensor> *result_ptr) {
PADDLE_ENFORCE("windows didn't supported recordio!.");
return true;
}
#endif // _WIN32
std::vector<LoDTensor> LoDTensor::SplitLoDTensor( std::vector<LoDTensor> LoDTensor::SplitLoDTensor(
const std::vector<platform::Place> places) const { const std::vector<platform::Place> places) const {
check_memory_size(); check_memory_size();
......
paddle/fluid/framework/lod_tensor_test.cc
浏览文件 @ be04d99f
...@@ -274,7 +274,6 @@ TEST(LoD, ConvertToOffsetBasedLoD) { ...@@ -274,7 +274,6 @@ TEST(LoD, ConvertToOffsetBasedLoD) {
EXPECT_EQ(offset_lod, expected); EXPECT_EQ(offset_lod, expected);
} }
#if !defined(_WIN32)
template <typename T> template <typename T>
static void TestRecordIO() { static void TestRecordIO() {
LoDTensor tensor; LoDTensor tensor;
...@@ -321,7 +320,6 @@ TEST(LoDTensor, RecordIO) { ...@@ -321,7 +320,6 @@ TEST(LoDTensor, RecordIO) {
TestRecordIO<float>(); TestRecordIO<float>();
TestRecordIO<double>(); TestRecordIO<double>();
} }
#endif // !defined(_WIN32)
} // namespace framework } // namespace framework
} // namespace paddle } // namespace paddle
paddle/fluid/framework/operator.cc
浏览文件 @ be04d99f
...@@ -149,17 +149,14 @@ void OperatorBase::Run(const Scope& scope, const platform::Place& place) { ...@@ -149,17 +149,14 @@ void OperatorBase::Run(const Scope& scope, const platform::Place& place) {
#endif #endif
} }
// The profile has a process-wide mutex, results in serious performance issue // The profile has a process-wide mutex, results in serious performance issue
// in concurrency scenerio. Here use an `if` to fix this issue. // in concurrency scenerio. Here use an `if` to fix this issue.
// Please not remove the `if`, ask @Superjomn if there are any concern. // Please not remove the `if`, ask @Superjomn if there are any concern.
#ifndef _WIN32
if (platform::IsProfileEnabled()) { if (platform::IsProfileEnabled()) {
platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance(); platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
platform::RecordEvent record_event(Type(), pool.Get(place)); platform::RecordEvent record_event(Type(), pool.Get(place));
RunImpl(scope, place); RunImpl(scope, place);
} else // NOLINT } else {
#endif
{
RunImpl(scope, place); RunImpl(scope, place);
} }
VLOG(3) << place << " " << DebugStringEx(&scope); VLOG(3) << place << " " << DebugStringEx(&scope);
......
paddle/fluid/inference/analysis/analyzer_tester.cc
浏览文件 @ be04d99f
...@@ -19,6 +19,7 @@ ...@@ -19,6 +19,7 @@
#include "paddle/fluid/inference/analysis/ut_helper.h" #include "paddle/fluid/inference/analysis/ut_helper.h"
#include "paddle/fluid/inference/api/paddle_inference_api.h" #include "paddle/fluid/inference/api/paddle_inference_api.h"
#include "paddle/fluid/inference/api/paddle_inference_pass.h" #include "paddle/fluid/inference/api/paddle_inference_pass.h"
#include "paddle/fluid/platform/port.h"
namespace paddle { namespace paddle {
namespace inference { namespace inference {
...@@ -75,7 +76,7 @@ void TestWord2vecPrediction(const std::string& model_path) { ...@@ -75,7 +76,7 @@ void TestWord2vecPrediction(const std::string& model_path) {
0.000932706}; 0.000932706};
const size_t num_elements = outputs.front().data.length() / sizeof(float); const size_t num_elements = outputs.front().data.length() / sizeof(float);
// The outputs' buffers are in CPU memory. // The outputs' buffers are in CPU memory.
for (size_t i = 0; i < std::min(5UL, num_elements); i++) { for (size_t i = 0; i < std::min((size_t)5UL, num_elements); i++) {
LOG(INFO) << "data: " LOG(INFO) << "data: "
<< static_cast<float*>(outputs.front().data.data())[i]; << static_cast<float*>(outputs.front().data.data())[i];
PADDLE_ENFORCE(static_cast<float*>(outputs.front().data.data())[i], PADDLE_ENFORCE(static_cast<float*>(outputs.front().data.data())[i],
......
paddle/fluid/inference/api/analysis_predictor.cc
浏览文件 @ be04d99f
...@@ -56,7 +56,6 @@ bool AnalysisPredictor::Init( ...@@ -56,7 +56,6 @@ bool AnalysisPredictor::Init(
const std::shared_ptr<framework::Scope> &parent_scope, const std::shared_ptr<framework::Scope> &parent_scope,
const std::shared_ptr<framework::ProgramDesc> &program) { const std::shared_ptr<framework::ProgramDesc> &program) {
VLOG(3) << "Predictor::init()"; VLOG(3) << "Predictor::init()";
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
LOG(WARNING) << "Profiler is actived, might affect the performance"; LOG(WARNING) << "Profiler is actived, might affect the performance";
LOG(INFO) << "You can turn off by set gflags '-profile false'"; LOG(INFO) << "You can turn off by set gflags '-profile false'";
...@@ -64,7 +63,6 @@ bool AnalysisPredictor::Init( ...@@ -64,7 +63,6 @@ bool AnalysisPredictor::Init(
: platform::ProfilerState::kCPU; : platform::ProfilerState::kCPU;
platform::EnableProfiler(tracking_device); platform::EnableProfiler(tracking_device);
} }
#endif
// no matter with or without MKLDNN // no matter with or without MKLDNN
paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads()); paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads());
...@@ -520,12 +518,10 @@ bool AnalysisPredictor::LoadParameters() { ...@@ -520,12 +518,10 @@ bool AnalysisPredictor::LoadParameters() {
} }
AnalysisPredictor::~AnalysisPredictor() { AnalysisPredictor::~AnalysisPredictor() {
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
platform::DisableProfiler(platform::EventSortingKey::kTotal, platform::DisableProfiler(platform::EventSortingKey::kTotal,
"./profile.log"); "./profile.log");
} }
#endif
if (sub_scope_) { if (sub_scope_) {
scope_->DeleteScope(sub_scope_); scope_->DeleteScope(sub_scope_);
} }
......
paddle/fluid/inference/api/api_impl.cc
浏览文件 @ be04d99f
...@@ -64,7 +64,6 @@ void NativePaddlePredictor::PrepareFeedFetch() { ...@@ -64,7 +64,6 @@ void NativePaddlePredictor::PrepareFeedFetch() {
bool NativePaddlePredictor::Init( bool NativePaddlePredictor::Init(
std::shared_ptr<framework::Scope> parent_scope) { std::shared_ptr<framework::Scope> parent_scope) {
VLOG(3) << "Predictor::init()"; VLOG(3) << "Predictor::init()";
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
LOG(WARNING) << "Profiler is actived, might affect the performance"; LOG(WARNING) << "Profiler is actived, might affect the performance";
LOG(INFO) << "You can turn off by set gflags '-profile false'"; LOG(INFO) << "You can turn off by set gflags '-profile false'";
...@@ -73,7 +72,6 @@ bool NativePaddlePredictor::Init( ...@@ -73,7 +72,6 @@ bool NativePaddlePredictor::Init(
: platform::ProfilerState::kCPU; : platform::ProfilerState::kCPU;
platform::EnableProfiler(tracking_device); platform::EnableProfiler(tracking_device);
} }
#endif
// no matter with or without MKLDNN // no matter with or without MKLDNN
paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads()); paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads());
...@@ -121,12 +119,10 @@ bool NativePaddlePredictor::Init( ...@@ -121,12 +119,10 @@ bool NativePaddlePredictor::Init(
} }
NativePaddlePredictor::~NativePaddlePredictor() { NativePaddlePredictor::~NativePaddlePredictor() {
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
platform::DisableProfiler(platform::EventSortingKey::kTotal, platform::DisableProfiler(platform::EventSortingKey::kTotal,
"./profile.log"); "./profile.log");
} }
#endif
if (sub_scope_) { if (sub_scope_) {
scope_->DeleteScope(sub_scope_); scope_->DeleteScope(sub_scope_);
} }
......
paddle/fluid/inference/api/helper.h
浏览文件 @ be04d99f
...@@ -15,10 +15,6 @@ ...@@ -15,10 +15,6 @@
#pragma once #pragma once
#include <glog/logging.h> #include <glog/logging.h>
#if !defined(_WIN32)
#include <sys/time.h>
#else
#endif
#include <algorithm> #include <algorithm>
#include <chrono> // NOLINT #include <chrono> // NOLINT
...@@ -28,6 +24,7 @@ ...@@ -28,6 +24,7 @@
#include <string> #include <string>
#include <vector> #include <vector>
#include "paddle/fluid/inference/api/paddle_inference_api.h" #include "paddle/fluid/inference/api/paddle_inference_api.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/string/printf.h" #include "paddle/fluid/string/printf.h"
namespace paddle { namespace paddle {
......
paddle/fluid/inference/tests/api/anakin_rnn1_tester.cc
浏览文件 @ be04d99f
...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and ...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include <gflags/gflags.h> #include <gflags/gflags.h>
#include <sys/time.h>
#include <time.h> #include <time.h>
#include <algorithm> #include <algorithm>
#include <fstream> #include <fstream>
......
paddle/fluid/inference/tests/api/tester_helper.h
浏览文件 @ be04d99f
...@@ -178,11 +178,9 @@ void TestOneThreadPrediction( ...@@ -178,11 +178,9 @@ void TestOneThreadPrediction(
warmup_timer.tic(); warmup_timer.tic();
predictor->Run(inputs[0], outputs, batch_size); predictor->Run(inputs[0], outputs, batch_size);
PrintTime(batch_size, 1, 1, 0, warmup_timer.toc(), 1); PrintTime(batch_size, 1, 1, 0, warmup_timer.toc(), 1);
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
paddle::platform::ResetProfiler(); paddle::platform::ResetProfiler();
} }
#endif
} }
LOG(INFO) << "Run " << num_times << " times..."; LOG(INFO) << "Run " << num_times << " times...";
...@@ -232,11 +230,9 @@ void TestMultiThreadPrediction( ...@@ -232,11 +230,9 @@ void TestMultiThreadPrediction(
warmup_timer.tic(); warmup_timer.tic();
predictor->Run(inputs[0], outputs, batch_size); predictor->Run(inputs[0], outputs, batch_size);
PrintTime(batch_size, 1, num_threads, tid, warmup_timer.toc(), 1); PrintTime(batch_size, 1, num_threads, tid, warmup_timer.toc(), 1);
#if !defined(_WIN32)
if (FLAGS_profile) { if (FLAGS_profile) {
paddle::platform::ResetProfiler(); paddle::platform::ResetProfiler();
} }
#endif
} }
LOG(INFO) << "Thread " << tid << " run " << num_times << " times..."; LOG(INFO) << "Thread " << tid << " run " << num_times << " times...";
......
paddle/fluid/inference/tests/book/test_inference_nlp.cc
浏览文件 @ be04d99f
...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include <sys/time.h>
#include <time.h> #include <time.h>
#include <fstream> #include <fstream>
#include <thread> // NOLINT #include <thread> // NOLINT
......
paddle/fluid/inference/tests/test_helper.h
浏览文件 @ be04d99f
...@@ -20,6 +20,7 @@ limitations under the License. */ ...@@ -20,6 +20,7 @@ limitations under the License. */
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/inference/io.h" #include "paddle/fluid/inference/io.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/platform/profiler.h" #include "paddle/fluid/platform/profiler.h"
DECLARE_bool(use_mkldnn); DECLARE_bool(use_mkldnn);
......
paddle/fluid/operators/beam_search_op_test.cc
浏览文件 @ be04d99f
...@@ -46,7 +46,7 @@ void CreateInput(LoDTensor* ids, LoDTensor* scores) { ...@@ -46,7 +46,7 @@ void CreateInput(LoDTensor* ids, LoDTensor* scores) {
auto* scores_data = scores->mutable_data<float>(place); auto* scores_data = scores->mutable_data<float>(place);
vector<int64_t> _ids({4, 2, 5, 2, 1, 3, 3, 5, 2, 8, 2, 1}); vector<int64_t> _ids({4, 2, 5, 2, 1, 3, 3, 5, 2, 8, 2, 1});
vector<float> _scores( vector<float> _scores(
{0.5, 0.3, 0.2, 0.6, 0.3, 0.1, 0.9, 0.5, 0.1, 0.7, 0.5, 0.1}); {0.5f, 0.3f, 0.2f, 0.6f, 0.3f, 0.1f, 0.9f, 0.5f, 0.1f, 0.7f, 0.5f, 0.1f});
for (int i = 0; i < 12; i++) { for (int i = 0; i < 12; i++) {
ids_data[i] = _ids[i]; ids_data[i] = _ids[i];
...@@ -80,7 +80,7 @@ TEST(DISABLED_beam_search_op, run) { ...@@ -80,7 +80,7 @@ TEST(DISABLED_beam_search_op, run) {
ASSERT_EQ(sids.lod(), sscores.lod()); ASSERT_EQ(sids.lod(), sscores.lod());
vector<int> tids({4, 2, 3, 8}); vector<int> tids({4, 2, 3, 8});
vector<float> tscores({0.5, 0.6, 0.9, 0.7}); vector<float> tscores({0.5f, 0.6f, 0.9f, 0.7f});
for (int i = 0; i < 4; i++) { for (int i = 0; i < 4; i++) {
ASSERT_EQ(tids[i], sids.data<int64_t>()[i]); ASSERT_EQ(tids[i], sids.data<int64_t>()[i]);
......
paddle/fluid/operators/distributed/grpc_client.cc
浏览文件 @ be04d99f
...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include <sys/time.h>
#include <limits> #include <limits>
#include "glog/logging.h" // For VLOG #include "glog/logging.h" // For VLOG
...@@ -20,6 +19,7 @@ limitations under the License. */ ...@@ -20,6 +19,7 @@ limitations under the License. */
#include "paddle/fluid/operators/distributed/grpc_client.h" #include "paddle/fluid/operators/distributed/grpc_client.h"
#include "paddle/fluid/operators/distributed/grpc_serde.h" #include "paddle/fluid/operators/distributed/grpc_serde.h"
#include "paddle/fluid/operators/distributed/request_handler.h" #include "paddle/fluid/operators/distributed/request_handler.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/platform/profiler.h" #include "paddle/fluid/platform/profiler.h"
DECLARE_bool(rpc_disable_reuse_port); DECLARE_bool(rpc_disable_reuse_port);
......
paddle/fluid/operators/distributed/grpc_serde.cc
浏览文件 @ be04d99f
...@@ -15,7 +15,6 @@ limitations under the License. */ ...@@ -15,7 +15,6 @@ limitations under the License. */
#ifdef PADDLE_WITH_CUDA #ifdef PADDLE_WITH_CUDA
#include <nccl.h> #include <nccl.h>
#endif #endif
#include <sys/time.h>
#include <thread> // NOLINT #include <thread> // NOLINT
#include "google/protobuf/io/coded_stream.h" #include "google/protobuf/io/coded_stream.h"
...@@ -26,6 +25,7 @@ limitations under the License. */ ...@@ -26,6 +25,7 @@ limitations under the License. */
#include "paddle/fluid/operators/distributed/grpc_variable_response.h" #include "paddle/fluid/operators/distributed/grpc_variable_response.h"
#include "paddle/fluid/operators/distributed/proto_encoder_helper.h" #include "paddle/fluid/operators/distributed/proto_encoder_helper.h"
#include "paddle/fluid/operators/distributed/sendrecvop_utils.h" #include "paddle/fluid/operators/distributed/sendrecvop_utils.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/platform/profiler.h" #include "paddle/fluid/platform/profiler.h"
namespace paddle { namespace paddle {
......
paddle/fluid/operators/distributed/grpc_serde.h
浏览文件 @ be04d99f
...@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and ...@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#pragma once #pragma once
#include <sys/time.h>
#include <iostream> #include <iostream>
#include <string> #include <string>
#include <vector> #include <vector>
...@@ -25,6 +25,7 @@ limitations under the License. */ ...@@ -25,6 +25,7 @@ limitations under the License. */
#include "paddle/fluid/framework/tensor_util.h" #include "paddle/fluid/framework/tensor_util.h"
#include "paddle/fluid/framework/var_type.h" #include "paddle/fluid/framework/var_type.h"
#include "paddle/fluid/operators/distributed/sendrecvop_utils.h" #include "paddle/fluid/operators/distributed/sendrecvop_utils.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/operators/distributed/send_recv.grpc.pb.h" #include "paddle/fluid/operators/distributed/send_recv.grpc.pb.h"
#include "paddle/fluid/operators/distributed/send_recv.pb.h" #include "paddle/fluid/operators/distributed/send_recv.pb.h"
......
paddle/fluid/operators/distributed/sendrecvop_utils.cc
浏览文件 @ be04d99f
...@@ -15,12 +15,12 @@ limitations under the License. */ ...@@ -15,12 +15,12 @@ limitations under the License. */
#ifdef PADDLE_WITH_CUDA #ifdef PADDLE_WITH_CUDA
#include <nccl.h> #include <nccl.h>
#endif #endif
#include <sys/time.h>
#include <thread> // NOLINT #include <thread> // NOLINT
#include "paddle/fluid/framework/data_type.h" #include "paddle/fluid/framework/data_type.h"
#include "paddle/fluid/operators/distributed/sendrecvop_utils.h" #include "paddle/fluid/operators/distributed/sendrecvop_utils.h"
#include "paddle/fluid/operators/distributed/variable_response.h" #include "paddle/fluid/operators/distributed/variable_response.h"
#include "paddle/fluid/platform/port.h"
DEFINE_bool(rpc_disable_reuse_port, false, "Disable SO_REUSEPORT or not."); DEFINE_bool(rpc_disable_reuse_port, false, "Disable SO_REUSEPORT or not.");
......
paddle/fluid/operators/distributed/sendrecvop_utils.h
浏览文件 @ be04d99f
...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and ...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#pragma once #pragma once
#include <sys/time.h>
#include <iostream> #include <iostream>
#include <string> #include <string>
#include <vector> #include <vector>
...@@ -24,6 +23,7 @@ limitations under the License. */ ...@@ -24,6 +23,7 @@ limitations under the License. */
#include "paddle/fluid/framework/selected_rows.h" #include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/framework/tensor_util.h" #include "paddle/fluid/framework/tensor_util.h"
#include "paddle/fluid/framework/var_type.h" #include "paddle/fluid/framework/var_type.h"
#include "paddle/fluid/platform/port.h"
#include "paddle/fluid/operators/distributed/send_recv.pb.h" #include "paddle/fluid/operators/distributed/send_recv.pb.h"
......
paddle/fluid/operators/math/cpu_vec_test.cc
浏览文件 @ be04d99f
...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ...@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include <sys/time.h>
#include <cmath> #include <cmath>
#include <cstring> #include <cstring>
#include <random> #include <random>
...@@ -22,6 +21,7 @@ limitations under the License. */ ...@@ -22,6 +21,7 @@ limitations under the License. */
#include "gtest/gtest.h" #include "gtest/gtest.h"
#include "paddle/fluid/operators/math/cpu_vec.h" #include "paddle/fluid/operators/math/cpu_vec.h"
#include "paddle/fluid/platform/port.h"
inline double GetCurrentUS() { inline double GetCurrentUS() {
struct timeval time; struct timeval time;
......
paddle/fluid/operators/math/im2col_test.cc
浏览文件 @ be04d99f
...@@ -14,9 +14,9 @@ limitations under the License. */ ...@@ -14,9 +14,9 @@ limitations under the License. */
#include "paddle/fluid/operators/math/im2col.h" #include "paddle/fluid/operators/math/im2col.h"
#include <gtest/gtest.h> #include <gtest/gtest.h>
#include <sys/time.h>
#include <vector> #include <vector>
#include "paddle/fluid/operators/math/im2col_cfo_cpu.h" #include "paddle/fluid/operators/math/im2col_cfo_cpu.h"
#include "paddle/fluid/platform/port.h"
template <typename DeviceContext, typename Place> template <typename DeviceContext, typename Place>
void testIm2col() { void testIm2col() {
......
paddle/fluid/operators/math/jit_kernel_test.cc
浏览文件 @ be04d99f
...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and ...@@ -13,7 +13,6 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include "paddle/fluid/operators/math/jit_kernel.h" #include "paddle/fluid/operators/math/jit_kernel.h"
#include <sys/time.h>
#include <cmath> // for exp #include <cmath> // for exp
#include <cstring> // for memcpy #include <cstring> // for memcpy
#include <random> #include <random>
...@@ -22,6 +21,7 @@ limitations under the License. */ ...@@ -22,6 +21,7 @@ limitations under the License. */
#include "gflags/gflags.h" #include "gflags/gflags.h"
#include "glog/logging.h" #include "glog/logging.h"
#include "gtest/gtest.h" #include "gtest/gtest.h"
#include "paddle/fluid/platform/port.h"
#ifdef PADDLE_WITH_MKLML #ifdef PADDLE_WITH_MKLML
#include "paddle/fluid/platform/dynload/mklml.h" #include "paddle/fluid/platform/dynload/mklml.h"
......
paddle/fluid/platform/cudnn_helper.h
浏览文件 @ be04d99f
...@@ -62,7 +62,7 @@ inline const char* cudnnGetErrorString(cudnnStatus_t status) { ...@@ -62,7 +62,7 @@ inline const char* cudnnGetErrorString(cudnnStatus_t status) {
#define CUDNN_ENFORCE(condition) \ #define CUDNN_ENFORCE(condition) \
do { \ do { \
cudnnStatus_t status = condition; \ auto status = condition; \
if (UNLIKELY(status != CUDNN_STATUS_SUCCESS)) { \ if (UNLIKELY(status != CUDNN_STATUS_SUCCESS)) { \
PADDLE_THROW(::paddle::platform::cudnnGetErrorString(status)); \ PADDLE_THROW(::paddle::platform::cudnnGetErrorString(status)); \
} \ } \
......
paddle/fluid/platform/dynload/cudnn.h
浏览文件 @ be04d99f
...@@ -48,13 +48,13 @@ extern void EnforceCUDNNLoaded(const char* fn_name); ...@@ -48,13 +48,13 @@ extern void EnforceCUDNNLoaded(const char* fn_name);
#else #else
#define DECLARE_DYNAMIC_LOAD_CUDNN_WRAP(__name) \ #define DECLARE_DYNAMIC_LOAD_CUDNN_WRAP(__name) \
struct DynLoad__##__name { \ struct DynLoad__##__name { \
template <typename... Args> \ template <typename... Args> \
inline cudnnStatus_t operator()(Args... args) { \ inline auto operator()(Args... args) { \
return ::__name(args...); \ return ::__name(args...); \
} \ } \
}; \ }; \
extern DynLoad__##__name __name extern DynLoad__##__name __name
#endif #endif
......
paddle/fluid/platform/gpu_info.cc
浏览文件 @ be04d99f
...@@ -19,7 +19,16 @@ limitations under the License. */ ...@@ -19,7 +19,16 @@ limitations under the License. */
#include "gflags/gflags.h" #include "gflags/gflags.h"
#include "paddle/fluid/platform/enforce.h" #include "paddle/fluid/platform/enforce.h"
DEFINE_double(fraction_of_gpu_memory_to_use, 0.92, #ifndef _WIN32
const float fraction_of_gpu_memory_to_use = 0.92f;
#else
// fraction_of_gpu_memory_to_use cannot be too high on windows,
// since the win32 graphic sub-system can occupy some GPU memory
// which may lead to insufficient memory left for paddle
const float fraction_of_gpu_memory_to_use = 0.5f;
#endif
DEFINE_double(fraction_of_gpu_memory_to_use, fraction_of_gpu_memory_to_use,
"Allocate a trunk of gpu memory that is this fraction of the " "Allocate a trunk of gpu memory that is this fraction of the "
"total gpu memory size. Future memory usage will be allocated " "total gpu memory size. Future memory usage will be allocated "
"from the trunk. If the trunk doesn't have enough gpu memory, " "from the trunk. If the trunk doesn't have enough gpu memory, "
......
paddle/fluid/platform/stream_callback_manager.h
浏览文件 @ be04d99f
...@@ -14,11 +14,11 @@ ...@@ -14,11 +14,11 @@
#pragma once #pragma once
#include <ThreadPool.h>
#include <cuda.h> #include <cuda.h>
#include <cuda_runtime.h> #include <cuda_runtime.h>
#include <functional> #include <functional>
#include <memory> #include <memory>
#include "ThreadPool.h"
#include "paddle/fluid/platform/enforce.h" #include "paddle/fluid/platform/enforce.h"
namespace paddle { namespace paddle {
......
paddle/fluid/pybind/protobuf.cc
浏览文件 @ be04d99f
...@@ -30,11 +30,12 @@ namespace pybind11 { ...@@ -30,11 +30,12 @@ namespace pybind11 {
namespace detail { namespace detail {
// Can be replaced by a generic lambda in C++14 // Can be replaced by a generic lambda in C++14
struct variant_caster_visitor : public boost::static_visitor<handle> { struct __attribute__((visibility("hidden"))) paddle_variant_caster_visitor
: public boost::static_visitor<handle> {
return_value_policy policy; return_value_policy policy;
handle parent; handle parent;
variant_caster_visitor(return_value_policy policy, handle parent) paddle_variant_caster_visitor(return_value_policy policy, handle parent)
: policy(policy), parent(parent) {} : policy(policy), parent(parent) {}
template <class T> template <class T>
...@@ -44,10 +45,10 @@ struct variant_caster_visitor : public boost::static_visitor<handle> { ...@@ -44,10 +45,10 @@ struct variant_caster_visitor : public boost::static_visitor<handle> {
}; };
template <class Variant> template <class Variant>
struct variant_caster; struct paddle_variant_caster;
template <template <class...> class V, class... Ts> template <template <class...> class V, class... Ts>
struct variant_caster<V<Ts...>> { struct paddle_variant_caster<V<Ts...>> {
using Type = V<Ts...>; using Type = V<Ts...>;
template <typename T> template <typename T>
...@@ -90,7 +91,7 @@ struct variant_caster<V<Ts...>> { ...@@ -90,7 +91,7 @@ struct variant_caster<V<Ts...>> {
static handle cast(Type const &src, return_value_policy policy, static handle cast(Type const &src, return_value_policy policy,
handle parent) { handle parent) {
variant_caster_visitor visitor(policy, parent); paddle_variant_caster_visitor visitor(policy, parent);
return boost::apply_visitor(visitor, src); return boost::apply_visitor(visitor, src);
} }
...@@ -101,7 +102,7 @@ struct variant_caster<V<Ts...>> { ...@@ -101,7 +102,7 @@ struct variant_caster<V<Ts...>> {
// Add specialization for concrete variant type // Add specialization for concrete variant type
template <class... Args> template <class... Args>
struct type_caster<boost::variant<Args...>> struct type_caster<boost::variant<Args...>>
: variant_caster<boost::variant<Args...>> {}; : paddle_variant_caster<boost::variant<Args...>> {};
} // namespace detail } // namespace detail
} // namespace pybind11 } // namespace pybind11
......
paddle/fluid/pybind/pybind.cc
浏览文件 @ be04d99f
...@@ -86,12 +86,12 @@ bool IsCompiledWithDIST() { ...@@ -86,12 +86,12 @@ bool IsCompiledWithDIST() {
#endif #endif
} }
PYBIND11_PLUGIN(core) { PYBIND11_MODULE(core, m) {
// Not used, just make sure cpu_info.cc is linked. // Not used, just make sure cpu_info.cc is linked.
paddle::platform::CpuTotalPhysicalMemory(); paddle::platform::CpuTotalPhysicalMemory();
paddle::memory::allocation::UseAllocatorStrategyGFlag(); paddle::memory::allocation::UseAllocatorStrategyGFlag();
py::module m("core", "C++ core of PaddlePaddle"); m.doc() = "C++ core of PaddlePaddle";
// using framework in this function. Since it is inside a function, it will // using framework in this function. Since it is inside a function, it will
// not cause namespace pollution. // not cause namespace pollution.
...@@ -907,7 +907,6 @@ All parameter, weight, gradient are variables in Paddle. ...@@ -907,7 +907,6 @@ All parameter, weight, gradient are variables in Paddle.
}); });
BindRecordIOWriter(&m); BindRecordIOWriter(&m);
return m.ptr();
} }
} // namespace pybind } // namespace pybind
} // namespace paddle } // namespace paddle
paddle/fluid/pybind/tensor_py.h
浏览文件 @ be04d99f
...@@ -21,7 +21,6 @@ limitations under the License. */ ...@@ -21,7 +21,6 @@ limitations under the License. */
#include "paddle/fluid/memory/memcpy.h" #include "paddle/fluid/memory/memcpy.h"
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/float16.h" #include "paddle/fluid/platform/float16.h"
#include "pybind11/common.h"
#include "pybind11/numpy.h" #include "pybind11/numpy.h"
#include "pybind11/pybind11.h" #include "pybind11/pybind11.h"
......
paddle/legacy/cuda/include/hl_warpctc_wrap.h
浏览文件 @ be04d99f
...@@ -12,9 +12,9 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ...@@ -12,9 +12,9 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#ifndef _WIN32
#ifndef HL_WARPCTC_WRAP_H_ #ifndef HL_WARPCTC_WRAP_H_
#define HL_WARPCTC_WRAP_H_ #define HL_WARPCTC_WRAP_H_
#include "ctc.h" #include "ctc.h"
#include "hl_base.h" #include "hl_base.h"
...@@ -91,3 +91,4 @@ extern void hl_warpctc_get_workspace_size(const int* cpuLabelLengths, ...@@ -91,3 +91,4 @@ extern void hl_warpctc_get_workspace_size(const int* cpuLabelLengths,
size_t* bytes); size_t* bytes);
#endif // HL_WARPCTC_WRAP_H_ #endif // HL_WARPCTC_WRAP_H_
#endif
paddle/legacy/cuda/src/hl_cuda_device.cc
浏览文件 @ be04d99f
...@@ -132,11 +132,15 @@ inline pid_t gettid() { ...@@ -132,11 +132,15 @@ inline pid_t gettid() {
uint64_t tid; uint64_t tid;
pthread_threadid_np(NULL, &tid); pthread_threadid_np(NULL, &tid);
#else #else
#ifndef _WIN32
#ifndef __NR_gettid #ifndef __NR_gettid
#define __NR_gettid 224 #define __NR_gettid 224
#endif #endif
pid_t tid = syscall(__NR_gettid); pid_t tid = syscall(__NR_gettid);
#endif #endif
#else // _WIN32
pid_t tid = _getpid();
#endif // _WIN32
CHECK_NE((int)tid, -1); CHECK_NE((int)tid, -1);
return tid; return tid;
} }
......
paddle/legacy/utils/ThreadLocal.h
浏览文件 @ be04d99f
...@@ -14,10 +14,12 @@ limitations under the License. */ ...@@ -14,10 +14,12 @@ limitations under the License. */
#pragma once #pragma once
#ifndef _WIN32
#include <pthread.h> #include <pthread.h>
#include <sys/syscall.h> #include <sys/syscall.h>
#include <sys/types.h>
#include <unistd.h> #include <unistd.h>
#endif
#include <sys/types.h>
#include <map> #include <map>
#include <mutex> #include <mutex>
#include <random> #include <random>
......
paddle/legacy/utils/Util.h
浏览文件 @ be04d99f
...@@ -14,7 +14,9 @@ limitations under the License. */ ...@@ -14,7 +14,9 @@ limitations under the License. */
#pragma once #pragma once
#ifndef _WIN32
#include <sys/syscall.h> // for syscall() #include <sys/syscall.h> // for syscall()
#endif
#include <sys/types.h> #include <sys/types.h>
#include <algorithm> #include <algorithm>
#include <cmath> #include <cmath>
...@@ -40,6 +42,31 @@ inline int rand_r(unsigned int* seedp) { ...@@ -40,6 +42,31 @@ inline int rand_r(unsigned int* seedp) {
} }
#endif #endif
#ifdef _WIN32
#define NOMINMAX // msvc max/min macro conflict with std::min/max
#include <windows.h>
template <typename T>
inline int __builtin_clz(const T& value) {
DWORD leadning_zero = 0;
if (_BitScanReverse(&leadning_zero, value)) {
return static_cast<int>(sizeof(T) * 8 - leadning_zero);
} else {
return static_cast<int>(0);
}
}
inline int __builtin_clzl(const unsigned long& value) {
return __builtin_clz(value);
}
inline int __builtin_clzll(const unsigned long long& value) {
return __builtin_clz(value);
}
#define pid_t int
#endif
/** /**
* Loop over the elements in a container * Loop over the elements in a container
* TODO(yuyang18): It's this foreach useful? Why not use C++ 11 foreach, * TODO(yuyang18): It's this foreach useful? Why not use C++ 11 foreach,
......
paddle/scripts/paddle_build.sh
浏览文件 @ be04d99f
...@@ -149,7 +149,7 @@ function cmake_gen() { ...@@ -149,7 +149,7 @@ function cmake_gen() {
elif [ "$1" == "cp37-cp37m" ]; then elif [ "$1" == "cp37-cp37m" ]; then
export LD_LIBRARY_PATH=/opt/_internal/cpython-3.7.0/lib/:${LD_LIBRARY_PATH} export LD_LIBRARY_PATH=/opt/_internal/cpython-3.7.0/lib/:${LD_LIBRARY_PATH}
export PATH=/opt/_internal/cpython-3.7.0/bin/:${PATH} export PATH=/opt/_internal/cpython-3.7.0/bin/:${PATH}
export PYTHON_FLAGS="-DPYTHON_EXECUTABLE:FILEPATH=/opt/_internal/cpython-3.7.0/bin/python3 export PYTHON_FLAGS="-DPYTHON_EXECUTABLE:FILEPATH=/opt/_internal/cpython-3.7.0/bin/python3.7
-DPYTHON_INCLUDE_DIR:PATH=/opt/_internal/cpython-3.7.0/include/python3.7m -DPYTHON_INCLUDE_DIR:PATH=/opt/_internal/cpython-3.7.0/include/python3.7m
-DPYTHON_LIBRARIES:FILEPATH=/opt/_internal/cpython-3.7.0/lib/libpython3.so" -DPYTHON_LIBRARIES:FILEPATH=/opt/_internal/cpython-3.7.0/lib/libpython3.so"
fi fi
......
paddle/testing/CMakeLists.txt
浏览文件 @ be04d99f
...@@ -3,8 +3,10 @@ ...@@ -3,8 +3,10 @@
if(WITH_TESTING) if(WITH_TESTING)
add_library(paddle_test_main STATIC TestMain.cpp) add_library(paddle_test_main STATIC TestMain.cpp)
add_dependencies(paddle_test_main paddle_proto ${external_project_dependencies}) add_dependencies(paddle_test_main paddle_proto ${external_project_dependencies})
add_library(paddle_test_util STATIC TestUtil.cpp) if(NOT WIN32)
add_dependencies(paddle_test_util paddle_proto ${external_project_dependencies}) add_library(paddle_test_util STATIC TestUtil.cpp)
add_dependencies(paddle_test_util paddle_proto ${external_project_dependencies})
endif(NOT WIN32)
if(NOT MOBILE_INFERENCE) if(NOT MOBILE_INFERENCE)
cc_library(paddle_gtest_main SRCS paddle_gtest_main.cc DEPS device_context memory gtest gflags) cc_library(paddle_gtest_main SRCS paddle_gtest_main.cc DEPS device_context memory gtest gflags)
endif() endif()
......
python/paddle/fluid/metrics.py
浏览文件 @ be04d99f
...@@ -46,8 +46,8 @@ def _is_numpy_(var): ...@@ -46,8 +46,8 @@ def _is_numpy_(var):
def _is_number_(var): def _is_number_(var):
return isinstance(var, int) or isinstance(var, float) or (isinstance( return isinstance(var, int) or isinstance(var, np.int64) or isinstance(
var, np.ndarray) and var.shape == (1, )) var, float) or (isinstance(var, np.ndarray) and var.shape == (1, ))
def _is_number_or_matrix_(var): def _is_number_or_matrix_(var):
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或