Modify cmake for cross-compiling on arm architecture.

CMakeLists.txt

@@ -25,7 +25,6 @@ find_package(Git REQUIRED)
 find_package(Threads REQUIRED)
 
 include(system)
-include(simd)
 
 ################################ Configurations #######################################
 option(WITH_GPU         "Compile PaddlePaddle with NVIDIA GPU"          ${CUDA_FOUND})
@@ -65,6 +64,7 @@ include(external/openblas)  # download, build, install openblas
 include(external/swig)      # download, build, install swig
 include(external/warpctc)   # download, build, install warpctc
 
+include(simd)               # set simd flag
 include(package)            # set paddle packages
 include(cpplint)            # set paddle c++ style
 include(ccache)             # set ccache for compilation
@@ -74,7 +74,6 @@ include(flags)              # set paddle compile flags
 include(cudnn)              # set cudnn libraries
 include(version)            # set PADDLE_VERSION
 include(coveralls)          # set code coverage
-
 include(configure)          # add paddle env configuration
 
 include_directories("${PROJ_ROOT}")

cmake/configure.cmake

@@ -59,10 +59,5 @@ else()
     include_directories(${CUDA_TOOLKIT_INCLUDE})
 endif(NOT WITH_GPU)
 
-if(WITH_AVX)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${AVX_FLAG}")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${AVX_FLAG}")
-else(WITH_AVX)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SSE3_FLAG}")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SSE3_FLAG}")
-endif(WITH_AVX)
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SIMD_FLAG}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SIMD_FLAG}")

cmake/cudnn.cmake

+if(NOT WITH_GPU)
+    return()
+endif()
+
 set(CUDNN_ROOT "" CACHE PATH "CUDNN ROOT")
 find_path(CUDNN_INCLUDE_DIR cudnn.h
     PATHS ${CUDNN_ROOT} ${CUDNN_ROOT}/include

cmake/external/protobuf.cmake

@@ -12,15 +12,31 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-INCLUDE(ExternalProject)
+FIND_PACKAGE(Protobuf REQUIRED)
+IF(Protobuf_FOUND)
+    # Check protobuf library version.
+    execute_process(COMMAND ${PROTOBUF_PROTOC_EXECUTABLE} --version
+        OUTPUT_VARIABLE PROTOBUF_VERSION)
+    string(REPLACE "libprotoc " "" PROTOBUF_VERSION ${PROTOBUF_VERSION})
+    #MESSAGE("PROTOBUF_INCLUDE_DIR: " ${PROTOBUF_INCLUDE_DIR})
+    #MESSAGE("PROTOBUF_LITE_LIBRARY: " ${PROTOBUF_LITE_LIBRARY})
+    #MESSAGE("PROTOBUF_LIBRARY: " ${PROTOBUF_LIBRARY})
+    #MESSAGE("PROTOBUF_PROTOC_LIBRARY: " ${PROTOBUF_PROTOC_LIBRARY})
+    #MESSAGE("PROTOBUF_PROTOC_EXECUTABLE: " ${PROTOBUF_PROTOC_EXECUTABLE})
 
-SET(PROTOBUF_SOURCES_DIR ${THIRD_PARTY_PATH}/protobuf)
-SET(PROTOBUF_INSTALL_DIR ${THIRD_PARTY_PATH}/install/protobuf)
-SET(PROTOBUF_INCLUDE_DIR "${PROTOBUF_INSTALL_DIR}/include" CACHE PATH "protobuf include directory." FORCE)
+    set(PROTOBUF_3 OFF)
+    if (${PROTOBUF_VERSION} VERSION_GREATER "3.0.0" OR ${PROTOBUF_VERSION} VERSION_EQUAL "3.0.0")
+        set(PROTOBUF_3 ON)
+    endif()
+ELSE()
 
-INCLUDE_DIRECTORIES(${PROTOBUF_INCLUDE_DIR})
+    INCLUDE(ExternalProject)
+
+    SET(PROTOBUF_SOURCES_DIR ${THIRD_PARTY_PATH}/protobuf)
+    SET(PROTOBUF_INSTALL_DIR ${THIRD_PARTY_PATH}/install/protobuf)
+    SET(PROTOBUF_INCLUDE_DIR "${PROTOBUF_INSTALL_DIR}/include" CACHE PATH "protobuf include directory." FORCE)
 
-IF(WIN32)
+    IF(WIN32)
         SET(PROTOBUF_LITE_LIBRARY
             "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf-lite.lib" CACHE FILEPATH "protobuf lite library." FORCE)
         SET(PROTOBUF_LIBRARY
@@ -28,7 +44,7 @@ IF(WIN32)
         SET(PROTOBUF_PROTOC_LIBRARY
             "${PROTOBUF_INSTALL_DIR}/lib/libprotoc.lib" CACHE FILEPATH "protoc library." FORCE)
         SET(PROTOBUF_PROTOC_EXECUTABLE "${PROTOBUF_INSTALL_DIR}/bin/protoc.exe" CACHE FILEPATH "protobuf executable." FORCE)
-ELSE(WIN32)
+    ELSE(WIN32)
         SET(PROTOBUF_LITE_LIBRARY
             "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf-lite.a" CACHE FILEPATH "protobuf lite library." FORCE)
         SET(PROTOBUF_LIBRARY
@@ -36,9 +52,9 @@ ELSE(WIN32)
         SET(PROTOBUF_PROTOC_LIBRARY
             "${PROTOBUF_INSTALL_DIR}/lib/libprotoc.a" CACHE FILEPATH "protoc library." FORCE)
         SET(PROTOBUF_PROTOC_EXECUTABLE "${PROTOBUF_INSTALL_DIR}/bin/protoc" CACHE FILEPATH "protobuf executable." FORCE)
-ENDIF(WIN32)
+    ENDIF(WIN32)
 
-ExternalProject_Add(
+    ExternalProject_Add(
         protobuf
         ${EXTERNAL_PROJECT_LOG_ARGS}
         PREFIX          ${PROTOBUF_SOURCES_DIR}
@@ -54,6 +70,9 @@ ExternalProject_Add(
         -DCMAKE_BUILD_TYPE=Release
         -DCMAKE_INSTALL_PREFIX=${PROTOBUF_INSTALL_DIR}
         -DCMAKE_INSTALL_LIBDIR=lib
-)
+        )
 
-LIST(APPEND external_project_dependencies protobuf)
+    LIST(APPEND external_project_dependencies protobuf)
+ENDIF()
+
+INCLUDE_DIRECTORIES(${PROTOBUF_INCLUDE_DIR})

cmake/simd.cmake

@@ -2,6 +2,7 @@
 # so that PaddlePaddle can unleash the vectorization power of muticore.
 
 INCLUDE(CheckCXXSourceRuns)
+INCLUDE(CheckCXXSourceCompiles)
 
 IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
     set(MMX_FLAG "-mmmx")
@@ -9,12 +10,14 @@ IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID
     set(SSE3_FLAG "-msse3")
     SET(AVX_FLAG "-mavx")
     SET(AVX2_FLAG "-mavx2")
+    SET(NEON_FLAG "-mfloat-abi=softfp -mfpu=neon")
 ELSEIF(MSVC)
     set(MMX_FLAG "/arch:MMX")
     set(SSE2_FLAG "/arch:SSE2")
     set(SSE3_FLAG "/arch:SSE3")
     SET(AVX_FLAG "/arch:AVX")
     SET(AVX2_FLAG "/arch:AVX2")
+    #SET(NEON_FLAG "")
 ENDIF()
 
 # Check  MMX
@@ -73,4 +76,26 @@ int main()
     return 0;
 }" AVX2_FOUND)
 
-mark_as_advanced(MMX_FOUND SSE2_FOUND SSE3_FOUND AVX_FOUND AVX2_FOUND)
+# Check NEON
+set(CMAKE_REQUIRED_FLAGS ${NEON_FLAG})
+CHECK_CXX_SOURCE_COMPILES("
+#include <arm_neon.h>
+int main()
+{
+    float32x4_t a = {-1.0f, 2.0f, -3.0f, 4.0f};
+    float32x4_t b = {1.0f, 2.0f, 3.0f, 4.0f};
+    float32x4_t c = vaddq_f32(a, b);
+    return 0;
+}" NEON_FOUND)
+
+if(NEON_FOUND)
+    set(SIMD_FLAG ${NEON_FLAG})
+else(NEON_FOUND)
+    if(WITH_AVX)
+        set(SIMD_FLAG ${AVX_FLAG})
+    else(WITH_AVX)
+        set(SIMD_FLAG ${SSE3_FLAG})
+    endif(WITH_AVX)
+endif(NEON_FOUND)
+
+mark_as_advanced(MMX_FOUND SSE2_FOUND SSE3_FOUND AVX_FOUND AVX2_FOUND NEON_FOUND)