Merge branch 'develop' into enhance_hierachical_sigmod_op

Author: JiabinYang

.gitignore

@@ -1,9 +1,11 @@
+python/paddle/fluid/tests/unittests/reader_reset_test.recordio
 paddle/operators/check_t.save
 paddle/operators/check_tensor.ls
 paddle/operators/tensor.save
 python/paddle/v2/fluid/tests/book/image_classification_resnet.inference.model/
 python/paddle/v2/fluid/tests/book/image_classification_vgg.inference.model/
 python/paddle/v2/fluid/tests/book/label_semantic_roles.inference.model/
+paddle/fluid/operators/distributed/send_recv.proto
 *.DS_Store
 *.vs
 build/
@@ -28,4 +30,5 @@ third_party/
 build_*
 # clion workspace.
 cmake-build-*
+paddle/fluid/operators/distributed/send_recv.proto
 model_test

AUTHORS.md

@@ -42,7 +42,9 @@
 | QiJune | Jun Qi |
 | qingqing01 | Qing-Qing Dang |
 | reyoung | Yang Yu |
+| Sand3r- | Michal Gallus |
 | Superjom | Chun-Wei Yan |
+| tensor-tang | Jian Tang |
 | tianbingsz | Tian-Bing Xu |
 | tpatejko | Tomasz Patejko |
 | typhoonzero | Yi Wu |

CMakeLists.txt

@@ -26,6 +26,11 @@ message(STATUS "C compiler: ${CMAKE_C_COMPILER}, version: "
         "${CMAKE_C_COMPILER_ID} ${CMAKE_C_COMPILER_VERSION}")
 if(WIN32)
     set(CMAKE_STATIC_LIBRARY_PREFIX lib)
+    add_definitions("/DGOOGLE_GLOG_DLL_DECL=")
+    set(CMAKE_C_FLAGS_DEBUG   "${CMAKE_C_FLAGS_DEBUG} /bigobj /MTd")
+    set(CMAKE_C_FLAGS_RELEASE  "${CMAKE_C_FLAGS_RELEASE} /bigobj /MT")
+    set(CMAKE_CXX_FLAGS_DEBUG  "${CMAKE_CXX_FLAGS_DEBUG} /bigobj /MTd")
+    set(CMAKE_CXX_FLAGS_RELEASE   "${CMAKE_CXX_FLAGS_RELEASE} /bigobj /MT")
 endif(WIN32)
 
 if(NOT CMAKE_CROSSCOMPILING)
@@ -41,6 +46,7 @@ option(WITH_GPU         "Compile PaddlePaddle with NVIDIA GPU"          ${CUDA_F
 option(WITH_AMD_GPU     "Compile PaddlePaddle with AMD GPU"             OFF)
 option(WITH_AVX         "Compile PaddlePaddle with AVX intrinsics"      ${AVX_FOUND})
 option(WITH_MKL         "Compile PaddlePaddle with MKL support."        ${AVX_FOUND})
+option(WITH_NGRAPH      "Compile PaddlePaddle with nGraph support."     OFF)
 option(WITH_DSO         "Compile PaddlePaddle with dynamic linked CUDA" ON)
 option(WITH_TESTING     "Compile PaddlePaddle with unit testing"        OFF)
 option(WITH_SWIG_PY     "Compile PaddlePaddle with inference api"       ON)
@@ -62,10 +68,11 @@ option(WITH_DISTRIBUTE  "Compile with distributed support"              OFF)
 option(USE_EIGEN_FOR_BLAS   "Use matrix multiplication in Eigen"        OFF)
 option(EIGEN_USE_THREADS "Compile with multi-threaded Eigen"            OFF)
 option(WITH_ARM_FP16    "Use half precision support on armv8.2-a cpu"   OFF)
-option(WITH_FAST_BUNDLE_TEST    "Bundle tests that can be run in a single process together to reduce launch overhead"   OFF)
 option(WITH_CONTRIB     "Compile the third-party contributation"        OFF)
 option(REPLACE_ENFORCE_GLOG "Replace PADDLE_ENFORCE with glog/CHECK for better debug." OFF)
 option(WITH_ANAKIN      "Compile with Anakin library"                   OFF)
+option(ANAKIN_BUILD_FAT_BIN "Build anakin cuda fat-bin lib for all device plantform, ignored when WITH_ANAKIN=OFF" OFF)
+option(ANAKIN_BUILD_CROSS_PLANTFORM "Build anakin lib for any nvidia device plantform. ignored when WITH_ANAKIN=OFF" ON)
 option(WITH_GRPC     "Use grpc as the default rpc framework"            ${WITH_DISTRIBUTE})
 option(WITH_BRPC_RDMA     "Use brpc rdma as the rpc protocal"           OFF)
 option(ON_INFER         "Turn on inference optimization."               OFF)
@@ -104,6 +111,8 @@ if(ANDROID OR IOS)
         "Disable RDMA when cross-compiling for Android and iOS" FORCE)
     set(WITH_MKL OFF CACHE STRING
         "Disable MKL when cross-compiling for Android and iOS" FORCE)
+    set(WITH_NGRAPH OFF CACHE STRING
+        "Disable nGraph when cross-compiling for Android and iOS" FORCE)
     set(WITH_GOLANG OFF CACHE STRING
         "Disable golang when cross-compiling for Android and iOS" FORCE)
 
@@ -172,6 +181,7 @@ include(external/protobuf)  # download, build, install protobuf
 include(external/python)    # download, build, install python
 include(external/openblas)  # download, build, install openblas
 include(external/mkldnn)    # download, build, install mkldnn
+include(external/ngraph)    # download, build, install nGraph
 include(external/swig)      # download, build, install swig
 include(external/boost)     # download boost
 include(external/any)       # download libn::any
@@ -292,6 +302,14 @@ set(PADDLE_PYTHON_BUILD_DIR "${CMAKE_CURRENT_BINARY_DIR}/python/build")
 set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
 set(CMAKE_C_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
 
+if (ON_INFER)
+    message(STATUS "On inference mode, will take place some specific optimization.")
+    add_definitions(-DPADDLE_ON_INFERENCE)
+else()
+    #TODO(luotao), combine this warning with `make inference_lib_dist` command.
+    message(WARNING "On inference mode, will take place some specific optimization. Turn on the ON_INFER flag when building inference_lib only.")
+endif()
+
 add_subdirectory(paddle)
 if(WITH_PYTHON)
     add_subdirectory(python)
@@ -302,11 +320,3 @@ if(WITH_DOC)
     find_python_module(recommonmark REQUIRED)
     add_subdirectory(doc)
 endif()
-
-if (ON_INFER)
-    message(STATUS "On inference mode, will take place some specific optimization.")
-    add_definitions(-DPADDLE_ON_INFERENCE)
-else()
-    #TODO(luotao), combine this warning with `make inference_lib_dist` command.
-    message(WARNING "On inference mode, will take place some specific optimization. Turn on the ON_INFER flag when building inference_lib only.")
-endif()

cmake/configure.cmake

@@ -50,11 +50,7 @@ if(NOT WITH_PROFILER)
 endif(NOT WITH_PROFILER)
 
 if(NOT CMAKE_CROSSCOMPILING)
-    if(WITH_AVX AND AVX512F_FOUND)
-        set(SIMD_FLAG ${AVX512F_FLAG})
-    elseif(WITH_AVX AND AVX2_FOUND)
-        set(SIMD_FLAG ${AVX2_FLAG})
-    elseif(WITH_AVX AND AVX_FOUND)
+    if(WITH_AVX AND AVX_FOUND)
         set(SIMD_FLAG ${AVX_FLAG})
     elseif(SSE3_FOUND)
         set(SIMD_FLAG ${SSE3_FLAG})
@@ -222,3 +218,7 @@ endif(WITH_GRPC)
 if(WITH_BRPC_RDMA)
     add_definitions(-DPADDLE_WITH_BRPC_RDMA)
 endif(WITH_BRPC_RDMA)
+
+if(ON_INFER)
+    add_definitions(-DPADDLE_ON_INFERENCE)
+endif(ON_INFER)

cmake/cuda.cmake

@@ -157,6 +157,9 @@ list(APPEND EXTERNAL_LIBS ${CUDA_LIBRARIES} ${CUDA_rt_LIBRARY})
 if(NOT WITH_DSO)
     # TODO(panyx0718): CUPTI only allows DSO?
     list(APPEND EXTERNAL_LIBS ${CUDNN_LIBRARY} ${CUPTI_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY} ${NCCL_LIBRARY})
+    if(WIN32)
+      set_property(GLOBAL PROPERTY CUDA_MODULES ${CUDNN_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY})
+    endif(WIN32)
 endif(NOT WITH_DSO)
 
 # setting nvcc arch flags
@@ -196,10 +199,12 @@ elseif(CMAKE_BUILD_TYPE  STREQUAL "MinSizeRel")
     list(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELEASE})
 endif()
 else(NOT WIN32)
-if(CMAKE_BUILD_TYPE STREQUAL "Release")
+if(CMAKE_BUILD_TYPE  STREQUAL "Debug")
+    list(APPEND CUDA_NVCC_FLAGS  "-g -G")
+elseif(CMAKE_BUILD_TYPE STREQUAL "Release")
   list(APPEND CUDA_NVCC_FLAGS "-O3 -DNDEBUG")
 else()
-  message(FATAL "Windows only support Release build now. Please set visual studio build type to Release, x64 build.")
+  message(FATAL "Windows only support Release or Debug build now. Please set visual studio build type to Release/Debug, x64 build.")
 endif()
 endif(NOT WIN32)
 

cmake/cudnn.cmake

@@ -2,7 +2,12 @@ if(NOT WITH_GPU)
     return()
 endif()
 
-set(CUDNN_ROOT "/usr" CACHE PATH "CUDNN ROOT")
+if(WIN32)
+    set(CUDNN_ROOT ${CUDA_TOOLKIT_ROOT_DIR})
+else(WIN32)
+    set(CUDNN_ROOT "/usr" CACHE PATH "CUDNN ROOT")
+endif(WIN32)
+
 find_path(CUDNN_INCLUDE_DIR cudnn.h
     PATHS ${CUDNN_ROOT} ${CUDNN_ROOT}/include
     $ENV{CUDNN_ROOT} $ENV{CUDNN_ROOT}/include ${CUDA_TOOLKIT_INCLUDE}

cmake/external/anakin.cmake

@@ -58,19 +58,21 @@ ExternalProject_Add(
                         -DPROTOBUF_ROOT=${THIRD_PARTY_PATH}/install/protobuf
                         -DMKLML_ROOT=${THIRD_PARTY_PATH}/install/mklml
                         -DENABLE_OP_TIMER=${ANAKIN_ENABLE_OP_TIMER}
+                        -DBUILD_FAT_BIN=${ANAKIN_BUILD_FAT_BIN}
+                        -DBUILD_CROSS_PLANTFORM=${ANAKIN_BUILD_CROSS_PLANTFORM}
                         ${EXTERNAL_OPTIONAL_ARGS}
     CMAKE_CACHE_ARGS    -DCMAKE_INSTALL_PREFIX:PATH=${ANAKIN_INSTALL_DIR}
 )
 
 message(STATUS "Anakin for inference is enabled")
 message(STATUS "Anakin is set INCLUDE:${ANAKIN_INCLUDE} LIBRARY:${ANAKIN_LIBRARY}")
-
+add_dependencies(extern_anakin protobuf mklml)
 add_library(anakin_shared SHARED IMPORTED GLOBAL)
 set_property(TARGET anakin_shared PROPERTY IMPORTED_LOCATION ${ANAKIN_SHARED_LIB})
-add_dependencies(anakin_shared extern_anakin protobuf mklml)
+add_dependencies(anakin_shared extern_anakin)
 
 add_library(anakin_saber SHARED IMPORTED GLOBAL)
 set_property(TARGET anakin_saber PROPERTY IMPORTED_LOCATION ${ANAKIN_SABER_LIB})
-add_dependencies(anakin_saber extern_anakin protobuf mklml)
+add_dependencies(anakin_saber extern_anakin)
 
 list(APPEND external_project_dependencies anakin_shared anakin_saber)

cmake/external/boost.cmake

@@ -28,34 +28,28 @@ if((NOT DEFINED BOOST_TAR) OR (NOT DEFINED BOOST_URL))
     set(BOOST_TAR "boost_1_41_0" CACHE STRING "" FORCE)
     set(BOOST_URL "http://paddlepaddledeps.cdn.bcebos.com/${BOOST_TAR}.tar.gz" CACHE STRING "" FORCE)
 endif()
-IF (WIN32)
-    MESSAGE(WARNING, "In windows, boost can not be downloaded automaticlly, please build it manually and put it at " ${THIRD_PARTY_PATH}install/boost)
-else()
-    MESSAGE(STATUS "BOOST_TAR: ${BOOST_TAR}, BOOST_URL: ${BOOST_URL}")
-ENDIF(WIN32)
+
+MESSAGE(STATUS "BOOST_TAR: ${BOOST_TAR}, BOOST_URL: ${BOOST_URL}")
 
 set(BOOST_SOURCES_DIR ${THIRD_PARTY_PATH}/boost)
 set(BOOST_DOWNLOAD_DIR  "${BOOST_SOURCES_DIR}/src/${BOOST_PROJECT}")
-set(BOOST_INCLUDE_DIR "${BOOST_DOWNLOAD_DIR}/${BOOST_TAR}" CACHE PATH "boost include directory." FORCE)
-set_directory_properties(PROPERTIES CLEAN_NO_CUSTOM 1)
 
+set(BOOST_INCLUDE_DIR "${BOOST_DOWNLOAD_DIR}" CACHE PATH "boost include directory." FORCE)
+set_directory_properties(PROPERTIES CLEAN_NO_CUSTOM 1)
 include_directories(${BOOST_INCLUDE_DIR})
 
-if (NOT WIN32)
 ExternalProject_Add(
     ${BOOST_PROJECT}
     ${EXTERNAL_PROJECT_LOG_ARGS}
     DOWNLOAD_DIR          ${BOOST_DOWNLOAD_DIR}
-    DOWNLOAD_COMMAND      wget --no-check-certificate ${BOOST_URL} -c -q -O ${BOOST_TAR}.tar.gz
-    && tar zxf ${BOOST_TAR}.tar.gz
+    URL      ${BOOST_URL}
     DOWNLOAD_NO_PROGRESS  1
     PREFIX                ${BOOST_SOURCES_DIR}
     CONFIGURE_COMMAND     ""
     BUILD_COMMAND         ""
     INSTALL_COMMAND       ""
     UPDATE_COMMAND        ""
-)
-endif(NOT WIN32)
+    )
 
 if (${CMAKE_VERSION} VERSION_LESS "3.3.0" OR NOT WIN32)
     set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/boost_dummy.c)

cmake/external/gflags.cmake

@@ -35,7 +35,12 @@ ExternalProject_Add(
     CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
                     -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
                     -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
+                    -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG}
                     -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG}
+                    -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE}
+                    -DBUILD_STATIC_LIBS=ON
                     -DCMAKE_INSTALL_PREFIX=${GFLAGS_INSTALL_DIR}
                     -DCMAKE_POSITION_INDEPENDENT_CODE=ON
                     -DBUILD_TESTING=OFF
@@ -48,8 +53,8 @@ ExternalProject_Add(
 IF(WIN32)
   IF(NOT EXISTS "${GFLAGS_INSTALL_DIR}/lib/libgflags.lib")
     add_custom_command(TARGET extern_gflags POST_BUILD
-    COMMAND cmake -E rename ${GFLAGS_INSTALL_DIR}/lib/gflags_static.lib ${GFLAGS_INSTALL_DIR}/lib/libgflags.lib
-  )
+            COMMAND cmake -E copy ${GFLAGS_INSTALL_DIR}/lib/gflags_static.lib ${GFLAGS_INSTALL_DIR}/lib/libgflags.lib
+            )
   ENDIF()
 ENDIF(WIN32)
 ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL)

cmake/external/glog.cmake

@@ -46,7 +46,11 @@ ExternalProject_Add(
     CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
                     -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
                     -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
+                    -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG}
                     -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG}
+                    -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE}
                     -DCMAKE_INSTALL_PREFIX=${GLOG_INSTALL_DIR}
                     -DCMAKE_INSTALL_LIBDIR=${GLOG_INSTALL_DIR}/lib
                     -DCMAKE_POSITION_INDEPENDENT_CODE=ON
@@ -63,7 +67,7 @@ ExternalProject_Add(
 IF(WIN32)
   IF(NOT EXISTS "${GLOG_INSTALL_DIR}/lib/libglog.lib")
     add_custom_command(TARGET extern_glog POST_BUILD
-    COMMAND cmake -E rename ${GLOG_INSTALL_DIR}/lib/glog.lib ${GLOG_INSTALL_DIR}/lib/libglog.lib
+    COMMAND cmake -E copy ${GLOG_INSTALL_DIR}/lib/glog.lib ${GLOG_INSTALL_DIR}/lib/libglog.lib
   )
   ENDIF()
 ENDIF(WIN32)