Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add cross-compiling support for arm architecture. #1698

Merged
merged 18 commits into from
Apr 11, 2017
Merged

Add cross-compiling support for arm architecture. #1698

Show file tree
Hide file tree
Changes from 13 commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
ccd3d0a
Modify cmake for cross-compiling on arm architecture.
Xreki Mar 8, 2017
717f755
Include arm_neon.h on arm platform.
Xreki Mar 9, 2017
4430fc9
Merge branch 'develop' into build_arm
Xreki Mar 14, 2017
ba11f5c
Modify compiling flags to support neon and add the definition of PADD…
Xreki Mar 23, 2017
6478483
Use internal gemm wrapper instead.
Xreki Mar 23, 2017
909cc6f
Implement paddle::str::to_string instead of calling std::to_string.
Xreki Mar 23, 2017
2a601e0
Set the simd-related kernels used under arm toolchains.
Xreki Mar 23, 2017
12e2402
Disable SpinLock and ThreadBarrier.
Xreki Mar 23, 2017
f261dc6
Merge branch 'develop' into build_arm
Xreki Mar 23, 2017
38fa74e
Fix cmake error of failing to find UINT64_MAX.
Xreki Mar 24, 2017
82f4ae7
Add alternative implementation of SpinLock and ThreadBarrier,
Xreki Mar 27, 2017
bb4a60d
Merge branch 'develop' into build_arm
Xreki Mar 27, 2017
1b85642
Use native cross-compiling support for Android of cmake.
Xreki Mar 29, 2017
7690e93
Deliver the compiler configurations of Paddle to external libraries.
Xreki Mar 29, 2017
ea5b265
Fix the setting of simd flags in cmake when there is no avx and sse3 …
Xreki Mar 29, 2017
3f187c3
Fix cmake broken when using lower version.
Xreki Mar 29, 2017
844eedd
Add the check of Android API level which must not be lower than 21.
Xreki Mar 29, 2017
2ae3dd0
Merge branch 'develop' into build_arm
Xreki Mar 31, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 24 additions & 7 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,19 +12,26 @@
# See the License for the specific language governing permissions and
# limitations under the License

cmake_minimum_required(VERSION 3.0)

project(paddle CXX C)

set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
set(PROJ_ROOT ${CMAKE_SOURCE_DIR})

include(system)

if(ANDROID)
cmake_minimum_required(VERSION 3.7)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个需要最低3.7？我这边是3.2.2的也是可以编译的。

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

你用的是编译方式2吧，如果使用编译方式1，cmake3.2.2会出现如下Warning，而且不会自动设置编译选项：

CMake Warning:
  Manually-specified variables were not used by the project:

    CMAKE_ANDROID_ARCH_ABI
    CMAKE_ANDROID_ARM_MODE
    CMAKE_ANDROID_ARM_NEON
    CMAKE_ANDROID_STANDALONE_TOOLCHAIN

这些系统cmake变量是在3.7版本后才加入的。（cmake-toolchains文档）

else()
cmake_minimum_required(VERSION 3.0)
endif()

project(paddle CXX C)

find_package(Sphinx)
find_package(CUDA QUIET)
if(NOT CMAKE_CROSSCOMPILING)
find_package(CUDA QUIET)
endif(NOT CMAKE_CROSSCOMPILING)
find_package(Git REQUIRED)
find_package(Threads REQUIRED)

include(system)
include(simd)

################################ Configurations #######################################
Expand All @@ -51,6 +58,17 @@ if(NOT CMAKE_BUILD_TYPE)
FORCE)
endif()

if(ANDROID)
set(WITH_GPU OFF CACHE STRING
"Disable GPU when cross-compiling for Android" FORCE)
set(WITH_AVX OFF CACHE STRING
"Disable AVX when cross-compiling for Android" FORCE)
set(WITH_PYTHON OFF CACHE STRING
"Disable PYTHON when cross-compiling for Android" FORCE)
set(WITH_RDMA OFF CACHE STRING
"Disable RDMA when cross-compiling for Android" FORCE)
endif(ANDROID)

set(THIRD_PARTY_PATH "${PROJ_ROOT}/third_party" CACHE STRING
"A path setting third party libraries download & build directories.")
########################################################################################
Expand All @@ -74,7 +92,6 @@ 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}")
Expand Down
28 changes: 20 additions & 8 deletions cmake/cblas.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,9 +19,9 @@ set(CBLAS_FOUND OFF)
set(INTEL_ROOT "/opt/intel" CACHE PATH "Folder contains intel libs")
set(MKL_ROOT ${INTEL_ROOT}/mkl CACHE PATH "Folder contains MKL")

find_path(MKL_INCLUDE_DIR mkl.h PATHS
find_path(MKL_INC_DIR mkl.h PATHS
${MKL_ROOT}/include)
find_path(MKL_INCLUDE_DIR mkl_lapacke.h PATHS
find_path(MKL_LAPACK_INC_DIR mkl_lapacke.h PATHS
${MKL_ROOT}/include)
find_library(MKL_CORE_LIB NAMES mkl_core PATHS
${MKL_ROOT}/lib
Expand All @@ -34,15 +34,19 @@ find_library(MKL_INTEL_LP64 NAMES mkl_intel_lp64 PATHS
${MKL_ROOT}/lib/intel64)


if(MKL_INCLUDE_DIR AND MKL_CORE_LIB AND MKL_SEQUENTIAL_LIB AND MKL_INTEL_LP64)
if(MKL_INC_DIR AND MKL_CORE_LIB AND MKL_SEQUENTIAL_LIB AND MKL_INTEL_LP64)
set(CBLAS_PROVIDER MKL)
set(CBLAS_INC_DIR ${MKL_INCLUDE_DIR})
set(CBLAS_INC_DIR ${MKL_INC_DIR})
set(CBLAS_LIBRARIES ${MKL_INTEL_LP64}
${MKL_SEQUENTIAL_LIB}
${MKL_CORE_LIB})
add_definitions(-DPADDLE_USE_MKL)
message(STATUS "Found MKL (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
set(CBLAS_FOUND ON)
if(${MKL_LAPACK_INC_DIR})
add_definitions(-DPADDLE_USE_LAPACK)
message(STATUS "Found lapack in MKL (include: ${MKL_LAPACK_INC_DIR})")
endif()
return() # return file.
endif()

Expand All @@ -68,13 +72,17 @@ find_library(ATLAS_CBLAS_LIB NAMES cblas libcblas.so.3
find_library(ATLAS_LIB NAMES lapack_atlas liblapack_atlas.so.3
PATHS ${ATLAS_LIB_SEARCH_PATHS})

if(ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_LIB)
if(ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_LIB AND NOT CBLAS_FOUND)
set(CBLAS_PROVIDER ATLAS)
set(CBLAS_INC_DIR ${ATLAS_INC_DIR} ${ATLAS_CLAPACK_INC_DIR})
set(CBLAS_INC_DIR ${ATLAS_INC_DIR})
set(CBLAS_LIBRARIES ${ATLAS_LIB} ${ATLAS_CBLAS_LIB})
add_definitions(-DPADDLE_USE_ATLAS)
message(STATUS "Found Atlas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
message(STATUS "Found ATLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
set(CBLAS_FOUND ON)
if(ATLAS_CLAPACK_INC_DIR)
add_definitions(-DPADDLE_USE_LAPACK)
message(STATUS "Found lapack in ATLAS (include: ${ATLAS_CLAPACK_INC_DIR})")
endif()
return()
endif()

Expand Down Expand Up @@ -103,8 +111,12 @@ if(OPENBLAS_INC_DIR AND OPENBLAS_LIB)
set(CBLAS_PROVIDER OPENBLAS)
set(CBLAS_INC_DIR ${OPENBLAS_INC_DIR})
set(CBLAS_LIBRARIES ${OPENBLAS_LIB})
message(STATUS "Found OpenBlas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
message(STATUS "Found OpenBLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
set(CBLAS_FOUND ON)
if(OPENBLAS_LAPACKE_INC_DIR)
add_definitions(-DPADDLE_USE_LAPACK)
message(STATUS "Found lapack in OpenBLAS (include: ${OPENBLAS_LAPACKE_INC_DIR})")
endif()
return()
endif()

Expand Down
23 changes: 11 additions & 12 deletions cmake/configure.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,14 @@ if(NOT WITH_PROFILER)
add_definitions(-DPADDLE_DISABLE_PROFILER)
endif(NOT WITH_PROFILER)

if(NOT CMAKE_CROSSCOMPILING)
if(WITH_AVX)
set(SIMD_FLAG ${AVX_FLAG})
else(WITH_AVX)
set(SIMD_FLAG ${SSE3_FLAG})
endif(WITH_AVX)
endif()

if(NOT WITH_GPU)
add_definitions(-DPADDLE_ONLY_CPU)
add_definitions(-DHPPL_STUB_FUNC)
Expand All @@ -48,21 +56,12 @@ else()
message(FATAL_ERROR "Paddle need cudnn to compile")
endif()

if(WITH_AVX)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${AVX_FLAG}")
else(WITH_AVX)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${SSE3_FLAG}")
endif(WITH_AVX)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${SIMD_FLAG}")

# Include cuda and cudnn
include_directories(${CUDNN_INCLUDE_DIR})
include_directories(${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}")
4 changes: 4 additions & 0 deletions cmake/cudnn.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,7 @@
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
Expand Down
2 changes: 2 additions & 0 deletions cmake/external/openblas.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,8 @@ IF(NOT ${CBLAS_FOUND})
"you need to set gfortran compiler: cmake .. -DCMAKE_Fortran_COMPILER=...")
ENDIF(NOT CMAKE_Fortran_COMPILER)

ADD_DEFINITIONS(-DPADDLE_USE_LAPACK)

ExternalProject_Add(
openblas
${EXTERNAL_PROJECT_LOG_ARGS}
Expand Down
8 changes: 4 additions & 4 deletions cmake/external/python.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -219,9 +219,9 @@ ELSE(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)

ENDIF(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)

INCLUDE_DIRECTORIES(${PYTHON_INCLUDE_DIR})
INCLUDE_DIRECTORIES(${PYTHON_NUMPY_INCLUDE_DIR})

IF(NOT WITH_PYTHON)
IF(WITH_PYTHON)
INCLUDE_DIRECTORIES(${PYTHON_INCLUDE_DIR})
INCLUDE_DIRECTORIES(${PYTHON_NUMPY_INCLUDE_DIR})
ELSE()
SET(PYTHON_LIBRARIES "")
ENDIF()
2 changes: 1 addition & 1 deletion cmake/external/zlib.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ SET(ZLIB_INCLUDE_DIR "${ZLIB_INSTALL_DIR}/include" CACHE PATH "zlib include dire
IF(WIN32)
SET(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/zlibstatic.lib" CACHE FILEPATH "zlib library." FORCE)
ELSE(WIN32)
set(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/libz.a" CACHE FILEPATH "zlib library." FORCE)
SET(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/libz.a" CACHE FILEPATH "zlib library." FORCE)
ENDIF(WIN32)

INCLUDE_DIRECTORIES(${ZLIB_INCLUDE_DIR})
Expand Down
14 changes: 13 additions & 1 deletion cmake/flags.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
include(CheckCXXCompilerFlag)
include(CheckCCompilerFlag)
include(CheckCXXSymbolExists)
include(CheckTypeSize)

function(CheckCompilerCXX11Flag)
if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
Expand All @@ -25,7 +26,7 @@ function(CheckCompilerCXX11Flag)
endfunction()

CheckCompilerCXX11Flag()
LIST(APPEND CMAKE_CXX_FLAGS -std=c++11)
string(APPEND CMAKE_CXX_FLAGS " -std=c++11")

# safe_set_flag
#
Expand Down Expand Up @@ -83,6 +84,17 @@ if(NOT UINT64_MAX_EXISTS)
endif()
endif()

SET(CMAKE_EXTRA_INCLUDE_FILES "pthread.h")
CHECK_TYPE_SIZE(pthread_spinlock_t SPINLOCK_FOUND)
CHECK_TYPE_SIZE(pthread_barrier_t BARRIER_FOUND)
if(SPINLOCK_FOUND)
add_definitions(-DPADDLE_USE_PTHREAD_SPINLOCK)
endif(SPINLOCK_FOUND)
if(BARRIER_FOUND)
add_definitions(-DPADDLE_USE_PTHREAD_BARRIER)
endif(BARRIER_FOUND)
SET(CMAKE_EXTRA_INCLUDE_FILES "")

# Common flags. the compiler flag used for C/C++ sources whenever release or debug
# Do not care if this flag is support for gcc.
set(COMMON_FLAGS
Expand Down
4 changes: 4 additions & 0 deletions cmake/simd.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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")
Expand All @@ -17,6 +18,8 @@ ELSEIF(MSVC)
SET(AVX2_FLAG "/arch:AVX2")
ENDIF()

set(CMAKE_REQUIRED_FLAGS_RETAINED ${CMAKE_REQUIRED_FLAGS})

# Check MMX
set(CMAKE_REQUIRED_FLAGS ${MMX_FLAG})
CHECK_CXX_SOURCE_RUNS("
Expand Down Expand Up @@ -73,4 +76,5 @@ int main()
return 0;
}" AVX2_FOUND)

set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_RETAINED})
mark_as_advanced(MMX_FOUND SSE2_FOUND SSE3_FOUND AVX_FOUND AVX2_FOUND)
6 changes: 6 additions & 0 deletions cmake/system.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,12 @@ MARK_AS_ADVANCED(HOST_SYSTEM CPU_CORES)
MESSAGE(STATUS "Found Paddle host system: ${HOST_SYSTEM}")
MESSAGE(STATUS "Found Paddle host system's CPU: ${CPU_CORES} cores")

IF(DEFINED CMAKE_SYSTEM_NAME)
IF(${CMAKE_SYSTEM_NAME} STREQUAL "Android")
SET(ANDROID TRUE)
ENDIF()
ENDIF()

# external dependencies log output
SET(EXTERNAL_PROJECT_LOG_ARGS
LOG_DOWNLOAD 0 # Wrap download in script to log output
Expand Down
4 changes: 4 additions & 0 deletions cmake/util.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,6 +90,10 @@ function(link_paddle_exe TARGET_NAME)
${RDMA_LD_FLAGS}
${RDMA_LIBS})

if(ANDROID)
target_link_libraries(${TARGET_NAME} log)
endif(ANDROID)

add_dependencies(${TARGET_NAME} ${external_project_dependencies})
endfunction()

Expand Down
4 changes: 4 additions & 0 deletions paddle/cuda/include/hl_cpu_matrix_kernel.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,11 @@ limitations under the License. */

#include <stdio.h>
#include "hl_base.h"
#if defined(__ARM_NEON__) || defined(__ARM_NEON)
#include "hl_neon_matrix_kernel.cuh"
#else
#include "hl_sse_matrix_kernel.cuh"
#endif

/**
* @brief cpu element wise unary operator.
Expand Down
2 changes: 2 additions & 0 deletions paddle/cuda/include/hl_matrix_base.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,8 @@ typedef BaseOp SSESquaredDiff;
typedef BaseOp SSEFirst;
typedef BaseOp SSESecond;
typedef BaseOp SSEClassificationError;
#elif defined(__ARM__NEON__) || defined(__ARM_NEON)
#include "hl_matrix_base_neon.cuh"
#else
#include "hl_matrix_base_sse.cuh"
#endif
Expand Down
Loading