Merge remote-tracking branch 'upstream/develop' into windows/build

CMakeLists.txt

@@ -38,6 +38,7 @@ if(NOT CMAKE_CROSSCOMPILING)
 endif(NOT CMAKE_CROSSCOMPILING)
 find_package(Git REQUIRED)
 find_package(Threads REQUIRED)
+
 include(simd)
 
 ################################ Configurations #######################################

cmake/cuda.cmake

@@ -172,21 +172,18 @@ set(CUDA_PROPAGATE_HOST_FLAGS OFF)
 
 # Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
 # So, don't set these flags here.
-
 if (NOT WIN32) # windows msvc2015 support c++11 natively. 
-# -std=c++11 -fPIC not recoginize by msvc
+# -std=c++11 -fPIC not recoginize by msvc, -Xcompiler will be added by cmake.
 list(APPEND CUDA_NVCC_FLAGS "-std=c++11")
-# in cuda9, suppress cuda warning on eigen with "-w"
+list(APPEND CUDA_NVCC_FLAGS "-Xcompiler -fPIC")
-list(APPEND CUDA_NVCC_FLAGS "-w" "-Xcompiler -fPIC")
-else(NOT WIN32)
-list(APPEND CUDA_NVCC_FLAGS "-w" "-Xcompiler -fPIC" "-Xcompiler /w")
 endif(NOT WIN32)
 
 if(WITH_FAST_MATH)
   # Make use of fast math library. https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html
   list(APPEND CUDA_NVCC_FLAGS "--use_fast_math")
-endif(WITH_FAST_MATH)
+endif()
-
+# in cuda9, suppress cuda warning on eigen 
+list(APPEND CUDA_NVCC_FLAGS "-w")
 # Set :expt-relaxed-constexpr to suppress Eigen warnings
 list(APPEND CUDA_NVCC_FLAGS "--expt-relaxed-constexpr")
 

cmake/cudnn.cmake

@@ -53,6 +53,7 @@ find_library(CUDNN_LIBRARY NAMES ${CUDNN_LIB_NAME} # libcudnn_static.a
           NO_DEFAULT_PATH
     DOC "Path to cuDNN library.")
 
+
 if(CUDNN_INCLUDE_DIR AND CUDNN_LIBRARY)
     set(CUDNN_FOUND ON)
 else()

cmake/external/boost.cmake

@@ -43,7 +43,7 @@ ExternalProject_Add(
     ${EXTERNAL_PROJECT_LOG_ARGS}
     DOWNLOAD_DIR          ${BOOST_DOWNLOAD_DIR}
     URL      ${BOOST_URL}
-    DOWNLOAD_NO_PROGRESS  0
+    DOWNLOAD_NO_PROGRESS  1
     PREFIX                ${BOOST_SOURCES_DIR}
     CONFIGURE_COMMAND     ""
     BUILD_COMMAND         ""

cmake/external/gflags.cmake

@@ -51,10 +51,6 @@ ExternalProject_Add(
                      -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                      -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
-
-ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL)
-SET_PROPERTY(TARGET gflags PROPERTY IMPORTED_LOCATION ${GFLAGS_LIBRARIES})
-ADD_DEPENDENCIES(gflags extern_gflags)
 IF(WIN32)
   IF(NOT EXISTS "${GFLAGS_INSTALL_DIR}/lib/libgflags.lib")
     add_custom_command(TARGET extern_gflags POST_BUILD
@@ -62,6 +58,9 @@ IF(WIN32)
             )
   ENDIF()
 ENDIF(WIN32)
+ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET gflags PROPERTY IMPORTED_LOCATION ${GFLAGS_LIBRARIES})
+ADD_DEPENDENCIES(gflags extern_gflags)
 
 LIST(APPEND external_project_dependencies gflags)
 

cmake/external/gtest.cmake

@@ -52,7 +52,6 @@ IF(WITH_TESTING)
                         -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
                         -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
                         -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-                        -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
                         -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
                         -DCMAKE_POSITION_INDEPENDENT_CODE=ON
                         -DBUILD_GMOCK=ON
@@ -72,5 +71,6 @@ IF(WITH_TESTING)
     ADD_LIBRARY(gtest_main STATIC IMPORTED GLOBAL)
     SET_PROPERTY(TARGET gtest_main PROPERTY IMPORTED_LOCATION ${GTEST_MAIN_LIBRARIES})
     ADD_DEPENDENCIES(gtest_main extern_gtest)
+
     LIST(APPEND external_project_dependencies gtest gtest_main)
 ENDIF(WITH_TESTING)

cmake/external/mkldnn.cmake

@@ -45,7 +45,7 @@ IF(${CBLAS_PROVIDER} STREQUAL "MKLML")
 ELSE()
     MESSAGE(FATAL_ERROR "Should enable MKLML when build MKLDNN")
 ENDIF()
-SET(MKLDNN_FLAG "-Wno-error=strict-overflow -Wno-error=unused-result")
+SET(MKLDNN_FLAG "-Wno-error=strict-overflow -Wno-error=unused-result -Wno-error=array-bounds")
 SET(MKLDNN_FLAG "${MKLDNN_FLAG} -Wno-unused-result -Wno-unused-value")
 SET(MKLDNN_CFLAG "${CMAKE_C_FLAGS} ${MKLDNN_FLAG}")
 SET(MKLDNN_CXXFLAG "${CMAKE_CXX_FLAGS} ${MKLDNN_FLAG}")
@@ -54,7 +54,7 @@ ExternalProject_Add(
     ${EXTERNAL_PROJECT_LOG_ARGS}
     DEPENDS             ${MKLDNN_DEPENDS}
     GIT_REPOSITORY      "https://github.com/01org/mkl-dnn.git"
-    GIT_TAG             "64e03a1939e0d526aa8e9f2e3f7dc0ad8d372944"
+    GIT_TAG             "21fb5f2af1dd14e132af4f1b79160977ee487818"
     PREFIX              ${MKLDNN_SOURCES_DIR}
     UPDATE_COMMAND      ""
     CMAKE_ARGS          -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}

cmake/external/openblas.cmake

@@ -149,7 +149,6 @@ INCLUDE_DIRECTORIES(${CBLAS_INC_DIR})
 # linear algebra libraries for cc_library(xxx SRCS xxx.c DEPS cblas)
 SET(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/cblas_dummy.c)
 FILE(WRITE ${dummyfile} "const char *dummy_cblas = \"${dummyfile}\";")
-
 ADD_LIBRARY(cblas STATIC ${dummyfile})
 
 IF("${CBLAS_PROVIDER}" STREQUAL "MKLML")

cmake/flags.cmake

@@ -144,14 +144,11 @@ set(GPU_COMMON_FLAGS
     -Wno-error=unused-function  # Warnings in Numpy Header.
     -Wno-error=array-bounds # Warnings in Eigen::array
 )
+
 else(NOT WIN32)
 set(COMMON_FLAGS
-    -fPIC
-    -fno-omit-frame-pointer
     "/w") #disable all warnings.
 set(GPU_COMMON_FLAGS
-    -fPIC
-    -fno-omit-frame-pointer
     "/w") #disable all warnings
 endif(NOT WIN32)
 
@@ -167,8 +164,8 @@ endif(APPLE)
 if(LINUX)
     set(GPU_COMMON_FLAGS
         -Wall
-        -Werror
         -Wextra
+        -Werror
         ${GPU_COMMON_FLAGS})
 endif(LINUX)
 

cmake/generic.cmake

@@ -238,7 +238,6 @@ function(cc_library TARGET_NAME)
       # add libxxx.lib prefix in windows
       set(${TARGET_NAME}_LIB_NAME "${CMAKE_STATIC_LIBRARY_PREFIX}${TARGET_NAME}${CMAKE_STATIC_LIBRARY_SUFFIX}" CACHE STRING "output library name for target ${TARGET_NAME}")
   endif(WIN32)
-
   if(cc_library_SRCS)
     if(cc_library_SHARED OR cc_library_shared) # build *.so
       add_library(${TARGET_NAME} SHARED ${cc_library_SRCS})
@@ -351,11 +350,7 @@ function(cc_test TARGET_NAME)
     set(multiValueArgs SRCS DEPS ARGS)
     cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     add_executable(${TARGET_NAME} ${cc_test_SRCS})
-    if(WIN32) # in windows deps. shlwapi library.
-    target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog shlwapi)
-    else(WIN32)
     target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
-    endif(WIN32)
     add_dependencies(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
     add_test(NAME ${TARGET_NAME}
              COMMAND ${TARGET_NAME} ${cc_test_ARGS}
@@ -426,11 +421,7 @@ function(nv_test TARGET_NAME)
     set(multiValueArgs SRCS DEPS)
     cmake_parse_arguments(nv_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     cuda_add_executable(${TARGET_NAME} ${nv_test_SRCS})
-    if(WIN32)
-    target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog shlwapi)
-    else(WIN32)
     target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
-    endif(WIN32)
     add_dependencies(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main lod_tensor memory gtest gflags glog)
     add_test(${TARGET_NAME} ${TARGET_NAME})
     if (nv_test_SERIAL)

cmake/inference_lib.cmake

@@ -55,7 +55,7 @@ function(copy TARGET)
                     COMMAND mkdir -p "${dst}"
                     COMMAND cp -r "${src}" "${dst}"
                     COMMENT "copying ${src} -> ${dst}")
-        endif(WIN32)
+        endif (WIN32) # not windows
     endforeach ()
 endfunction()
 

cmake/version.cmake

@@ -44,5 +44,5 @@ while ("${PADDLE_VERSION}" STREQUAL "")
   endif()
 endwhile()
 
-add_definitions(-DPADDLE_VERSION="${PADDLE_VERSION}")
+add_definitions(-DPADDLE_VERSION=${PADDLE_VERSION})
 message(STATUS "Paddle version is ${PADDLE_VERSION}")

doc/fluid/dev/contribute_to_paddle_cn.md

@@ -1 +0,0 @@
-../../v2/dev/contribute_to_paddle_cn.md

doc/fluid/dev/contribute_to_paddle_en.md

@@ -1 +0,0 @@
-../../v2/dev/contribute_to_paddle_en.md

doc/fluid/new_docs/advanced_usage/development/contribute_to_paddle.md

@@ -1 +0,0 @@
-../../../dev/contribute_to_paddle_cn.md

doc/fluid/new_docs/advanced_usage/development/cpu_profiling_cn.md

@@ -1 +0,0 @@
-../../../howto/optimization/cpu_profiling_cn.md

doc/fluid/new_docs/advanced_usage/development/host_memory_profiling_cn.md

@@ -1 +0,0 @@
-../../../howto/optimization/host_memory_profiling_cn.md

doc/fluid/new_docs/advanced_usage/development/new_op.md

@@ -1 +0,0 @@
-../../../dev/new_op_cn.md

doc/fluid/new_docs/advanced_usage/development/timeline_cn.md

@@ -1 +0,0 @@
-../../../howto/optimization/timeline_cn.md

paddle/fluid/API.spec

@@ -118,9 +118,10 @@ paddle.fluid.layers.label_smooth ArgSpec(args=['label', 'prior_dist', 'epsilon',
 paddle.fluid.layers.roi_pool ArgSpec(args=['input', 'rois', 'pooled_height', 'pooled_width', 'spatial_scale'], varargs=None, keywords=None, defaults=(1, 1, 1.0))
 paddle.fluid.layers.roi_align ArgSpec(args=['input', 'rois', 'pooled_height', 'pooled_width', 'spatial_scale', 'sampling_ratio', 'name'], varargs=None, keywords=None, defaults=(1, 1, 1.0, -1, None))
 paddle.fluid.layers.dice_loss ArgSpec(args=['input', 'label', 'epsilon'], varargs=None, keywords=None, defaults=(1e-05,))
-paddle.fluid.layers.image_resize ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'resample'], varargs=None, keywords=None, defaults=(None, None, None, 'BILINEAR'))
+paddle.fluid.layers.image_resize ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'resample', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, 'BILINEAR', None))
 paddle.fluid.layers.image_resize_short ArgSpec(args=['input', 'out_short_len', 'resample'], varargs=None, keywords=None, defaults=('BILINEAR',))
-paddle.fluid.layers.resize_bilinear ArgSpec(args=['input', 'out_shape', 'scale', 'name'], varargs=None, keywords=None, defaults=(None, None, None))
+paddle.fluid.layers.resize_bilinear ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, None))
+paddle.fluid.layers.resize_nearest ArgSpec(args=['input', 'out_shape', 'scale', 'name', 'actual_shape'], varargs=None, keywords=None, defaults=(None, None, None, None))
 paddle.fluid.layers.gather ArgSpec(args=['input', 'index'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.scatter ArgSpec(args=['input', 'index', 'updates', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.sequence_scatter ArgSpec(args=['input', 'index', 'updates', 'name'], varargs=None, keywords=None, defaults=(None,))

paddle/fluid/framework/executor.cc

@@ -12,8 +12,6 @@ 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 <algorithm>
-
 #include "paddle/fluid/framework/executor.h"
 
 #include "paddle/fluid/framework/feed_fetch_method.h"
@@ -48,7 +46,6 @@ ExecutorPrepareContext::~ExecutorPrepareContext() {
   VLOG(5) << "destroy ExecutorPrepareContext";
 }
 
-#ifndef _WIN32
 template <typename RefCntMap>
 static void DeleteUnusedTensors(const Scope& scope, const OperatorBase* op,
                                 GarbageCollector<Tensor>* gc,
@@ -83,7 +80,6 @@ static void DeleteUnusedTensors(const Scope& scope, const OperatorBase* op,
     gc->Add(erase_tensors);
   }
 }
-#endif
 
 Executor::Executor(const platform::Place& place) : place_(place) {}
 
@@ -371,7 +367,6 @@ void Executor::RunPreparedContext(ExecutorPrepareContext* ctx, Scope* scope,
     CreateVariables(ctx->prog_, local_scope, ctx->block_id_);
   }
 
-#ifndef _WIN32
   int64_t max_memory_size = GetEagerDeletionThreshold();
   std::unique_ptr<GarbageCollector<Tensor>> gc;
   // WhileOp would set keep_kids to false
@@ -413,16 +408,6 @@ void Executor::RunPreparedContext(ExecutorPrepareContext* ctx, Scope* scope,
   } else {
     platform::DeviceContextPool::Instance().Get(place_)->Wait();
   }
-#else   // WIN32
-  for (auto& op : ctx->ops_) {
-    op->Run(*local_scope, place_);
-    if (FLAGS_benchmark) {
-      VLOG(2) << "Memory used after operator " + op->Type() + " running: "
-              << memory::memory_usage(place_);
-    }
-  }
-  platform::DeviceContextPool::Instance().Get(place_)->Wait();
-#endif  // NOT WIN32
 
   if (local_scope != scope) {
     scope->DeleteScope(local_scope);

paddle/fluid/framework/executor.h

@@ -17,14 +17,12 @@ limitations under the License. */
 #include <map>
 #include <string>
 #include <vector>
+#include "paddle/fluid/framework/garbage_collector.h"
 #include "paddle/fluid/framework/op_info.h"
 #include "paddle/fluid/framework/program_desc.h"
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/platform/device_context.h"
-#ifndef _WIN32
-#include "paddle/fluid/framework/garbage_collector.h"
-#endif
 
 namespace paddle {
 namespace framework {

paddle/fluid/framework/ir/graph_pattern_detector.cc

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include <algorithm>
 #include <array>
 #include <string>
 #include <vector>

paddle/fluid/framework/ir/pass.h

@@ -21,7 +21,6 @@ limitations under the License. */
 #include "paddle/fluid/framework/ir/graph.h"
 #include "paddle/fluid/framework/ir/node.h"
 #include "paddle/fluid/framework/program_desc.h"
-#include "paddle/fluid/platform/port.h"
 #include "paddle/fluid/platform/variant.h"
 
 namespace paddle {
@@ -196,7 +195,6 @@ struct PassRegistrar : public Registrar {
                              __test_global_namespace_##uniq_name##__>::value, \
                 msg)
 
-#if !defined(_WIN32)
 // Register a new pass that can be applied on the IR.
 #define REGISTER_PASS(pass_type, pass_class)                          \
   STATIC_ASSERT_PASS_GLOBAL_NAMESPACE(                                \
@@ -242,7 +240,6 @@ struct PassRegistrar : public Registrar {
   static int UNUSED(use_pass_itself_##pass_type##_) = \
       TouchPassRegistrar_##pass_type()
 
-#endif  // !_WIN32
 }  // namespace ir
 }  // namespace framework
 }  // namespace paddle

paddle/fluid/framework/operator.cc

@@ -361,7 +361,7 @@ static bool VarIsTensor(const Variable& var) {
   return var.IsType<LoDTensor>() || var.IsType<SelectedRows>();
 }
 
-const Tensor* GetTensorFromVar(const Variable& var) {
+const Tensor* GetLoDTensorOrSelectedRowsValueFromVar(const Variable& var) {
   if (var.IsType<LoDTensor>()) {
     return static_cast<const Tensor*>(&(var.Get<LoDTensor>()));
   } else if (var.IsType<SelectedRows>()) {
@@ -372,7 +372,7 @@ const Tensor* GetTensorFromVar(const Variable& var) {
   }
 }
 
-static Tensor* GetMutableTensorFromVar(Variable* var) {
+Tensor* GetMutableLoDTensorOrSelectedRowsValueFromVar(Variable* var) {
   if (var->IsType<LoDTensor>()) {
     return var->GetMutable<LoDTensor>();
   } else if (var->IsType<SelectedRows>()) {
@@ -417,8 +417,7 @@ bool ExecutionContext::HasOutput(const std::string& name) const {
 
 template <>
 const Tensor* ExecutionContext::Input<Tensor>(const std::string& name) const {
-  auto* var = InputVar(name);
+  return Input<LoDTensor>(name);
-  return var == nullptr ? nullptr : GetTensorFromVar(*var);
 }
 
 template <>
@@ -428,17 +427,21 @@ const std::vector<const Tensor*> ExecutionContext::MultiInput<Tensor>(
   std::vector<const Tensor*> res;
   res.reserve(names.size());
   std::transform(names.begin(), names.end(), std::back_inserter(res),
-                 [&](const std::string& sub_name) {
+                 [&](const std::string& sub_name) -> const Tensor* {
                    auto var = scope_.FindVar(sub_name);
-                   return var == nullptr ? nullptr : GetTensorFromVar(*var);
+                   if (var == nullptr) return nullptr;
+                   PADDLE_ENFORCE(
+                       var->IsType<LoDTensor>(),
+                       "%s should be LoDTensor, but the received type is %s",
+                       sub_name, var->Type().name());
+                   return &(var->Get<LoDTensor>());
                  });
   return res;
 }
 
 template <>
 Tensor* ExecutionContext::Output<Tensor>(const std::string& name) const {
-  auto var = OutputVar(name);
+  return Output<LoDTensor>(name);
-  return var == nullptr ? nullptr : GetMutableTensorFromVar(var);
 }
 
 template <>
@@ -448,10 +451,14 @@ std::vector<Tensor*> ExecutionContext::MultiOutput<Tensor>(
   std::vector<Tensor*> res;
   res.reserve(names.size());
   std::transform(names.begin(), names.end(), std::back_inserter(res),
-                 [&](const std::string& sub_name) {
+                 [&](const std::string& sub_name) -> Tensor* {
                    auto var = scope_.FindVar(sub_name);
-                   return var == nullptr ? nullptr
+                   if (var == nullptr) return nullptr;
-                                         : GetMutableTensorFromVar(var);
+                   PADDLE_ENFORCE(
+                       var->IsType<LoDTensor>(),
+                       "%s should be LoDTensor, but the received type is %s",
+                       sub_name, var->Type().name());
+                   return var->GetMutable<LoDTensor>();
                  });
   return res;
 }
@@ -771,11 +778,12 @@ void OperatorWithKernel::TransferInplaceVarsBack(
     const Scope& transfer_scope) const {
   for (auto& var_name : inplace_vars) {
     VLOG(3) << "share inplace var " + var_name + " back to it's original scope";
-    auto* original_tensor = GetMutableTensorFromVar(scope.FindVar(var_name));
+    auto* original_tensor =
+        GetMutableLoDTensorOrSelectedRowsValueFromVar(scope.FindVar(var_name));
     auto* var = transfer_scope.FindVar(var_name);
     PADDLE_ENFORCE(var != nullptr, "The var[%s] should not be nullptr",
                    var_name);
-    auto* transformed_tensor = GetTensorFromVar(*var);
+    auto* transformed_tensor = GetLoDTensorOrSelectedRowsValueFromVar(*var);
     original_tensor->ShareDataWith(*transformed_tensor);
   }
 }
@@ -792,7 +800,7 @@ Scope* OperatorWithKernel::TryTransferData(
         continue;
       }
 
-      auto* tensor_in = GetTensorFromVar(*var);
+      auto* tensor_in = GetLoDTensorOrSelectedRowsValueFromVar(*var);
       if (!tensor_in->IsInitialized()) {
         continue;
       }