opencv/modules/core/CMakeLists.txt
Sayed Adel f2fe6f40c2 Merge pull request #15510 from seiko2plus:issue15506
* core: rework and optimize SIMD implementation of dotProd

  - add new universal intrinsics v_dotprod[int32], v_dotprod_expand[u&int8, u&int16, int32], v_cvt_f64(int64)
  - add a boolean param for all v_dotprod&_expand intrinsics that change the behavior of addition order between
    pairs in some platforms in order to reach the maximum optimization when the sum among all lanes is what only matters
  - fix clang build on ppc64le
  - support wide universal intrinsics for dotProd_32s
  - remove raw SIMD and activate universal intrinsics for dotProd_8
  - implement SIMD optimization for dotProd_s16&u16
  - extend performance test data types of dotprod
  - fix GCC VSX workaround of vec_mule and vec_mulo (in little-endian it must be swapped)
  - optimize v_mul_expand(int32) on VSX

* core: remove boolean param from v_dotprod&_expand and implement v_dotprod_fast&v_dotprod_expand_fast

  this changes made depend on "terfendail" review
2019-10-07 22:01:35 +03:00

138 lines
5.1 KiB
CMake

set(the_description "The Core Functionality")
ocv_add_dispatched_file(mathfuncs_core SSE2 AVX AVX2)
ocv_add_dispatched_file(stat SSE4_2 AVX2)
ocv_add_dispatched_file(arithm SSE2 SSE4_1 AVX2 VSX3)
ocv_add_dispatched_file(convert SSE2 AVX2 VSX3)
ocv_add_dispatched_file(convert_scale SSE2 AVX2)
ocv_add_dispatched_file(count_non_zero SSE2 AVX2)
ocv_add_dispatched_file(matmul SSE2 SSE4_1 AVX2 AVX512_SKX)
ocv_add_dispatched_file(mean SSE2 AVX2)
ocv_add_dispatched_file(merge SSE2 AVX2)
ocv_add_dispatched_file(split SSE2 AVX2)
ocv_add_dispatched_file(sum SSE2 AVX2)
# dispatching for accuracy tests
ocv_add_dispatched_file_force_all(test_intrin128 TEST SSE2 SSE3 SSSE3 SSE4_1 SSE4_2 AVX FP16 AVX2 AVX512_SKX)
ocv_add_dispatched_file_force_all(test_intrin256 TEST AVX2 AVX512_SKX)
ocv_add_dispatched_file_force_all(test_intrin512 TEST AVX512_SKX)
ocv_add_module(core
OPTIONAL opencv_cudev
WRAP java python js)
set(extra_libs "")
if(WINRT AND WINDOWS_STORE AND CMAKE_SYSTEM_VERSION MATCHES "8.0")
list(APPEND extra_libs ole32.lib)
endif()
if(HAVE_TBB)
list(APPEND extra_libs tbb)
endif()
if(DEFINED WINRT AND NOT DEFINED ENABLE_WINRT_MODE_NATIVE)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /ZW")
endif()
if(HAVE_CUDA)
ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wenum-compare -Wunused-function -Wshadow)
endif()
if(CV_TRACE AND HAVE_ITT AND BUILD_ITT)
add_definitions(-DOPENCV_WITH_ITT=1)
endif()
file(GLOB lib_cuda_hdrs
"${CMAKE_CURRENT_LIST_DIR}/include/opencv2/${name}/cuda/*.hpp"
"${CMAKE_CURRENT_LIST_DIR}/include/opencv2/${name}/cuda/*.h")
file(GLOB lib_cuda_hdrs_detail
"${CMAKE_CURRENT_LIST_DIR}/include/opencv2/${name}/cuda/detail/*.hpp"
"${CMAKE_CURRENT_LIST_DIR}/include/opencv2/${name}/cuda/detail/*.h")
file(GLOB_RECURSE module_opencl_hdrs
"${CMAKE_CURRENT_LIST_DIR}/include/opencv2/${name}/opencl/*")
source_group("Include\\Cuda Headers" FILES ${lib_cuda_hdrs})
source_group("Include\\Cuda Headers\\Detail" FILES ${lib_cuda_hdrs_detail})
source_group("Src" FILES "${OPENCV_MODULE_opencv_core_BINARY_DIR}/version_string.inc")
ocv_glob_module_sources(SOURCES "${OPENCV_MODULE_opencv_core_BINARY_DIR}/version_string.inc"
HEADERS ${module_opencl_hdrs} ${lib_cuda_hdrs} ${lib_cuda_hdrs_detail})
ocv_module_include_directories(${the_module} ${ZLIB_INCLUDE_DIRS} ${OPENCL_INCLUDE_DIRS})
if(ANDROID AND HAVE_CPUFEATURES)
ocv_append_source_file_compile_definitions(${CMAKE_CURRENT_SOURCE_DIR}/src/system.cpp "HAVE_CPUFEATURES=1")
ocv_module_include_directories(${CPUFEATURES_INCLUDE_DIRS})
endif()
if(ITT_INCLUDE_DIRS)
ocv_module_include_directories(${ITT_INCLUDE_DIRS})
endif()
if(HAVE_POSIX_MEMALIGN)
ocv_append_source_file_compile_definitions(${CMAKE_CURRENT_SOURCE_DIR}/src/alloc.cpp "HAVE_POSIX_MEMALIGN=1")
endif()
if(HAVE_MALLOC_H)
ocv_append_source_file_compile_definitions(${CMAKE_CURRENT_SOURCE_DIR}/src/alloc.cpp "HAVE_MALLOC_H=1")
endif()
if(HAVE_MEMALIGN)
ocv_append_source_file_compile_definitions(${CMAKE_CURRENT_SOURCE_DIR}/src/alloc.cpp "HAVE_MEMALIGN=1")
endif()
ocv_create_module(${extra_libs})
ocv_target_link_libraries(${the_module} LINK_PRIVATE
"${ZLIB_LIBRARIES}" "${OPENCL_LIBRARIES}" "${VA_LIBRARIES}"
"${LAPACK_LIBRARIES}" "${CPUFEATURES_LIBRARIES}" "${HALIDE_LIBRARIES}"
"${ITT_LIBRARIES}"
"${OPENCV_HAL_LINKER_LIBS}"
)
ocv_add_accuracy_tests()
ocv_add_perf_tests()
ocv_install_3rdparty_licenses(SoftFloat "${CMAKE_CURRENT_SOURCE_DIR}/3rdparty/SoftFloat/COPYING.txt")
# generate data (samples data) config file
set(OPENCV_DATA_CONFIG_FILE "${CMAKE_BINARY_DIR}/opencv_data_config.hpp")
set(OPENCV_DATA_CONFIG_STR "")
if(CMAKE_INSTALL_PREFIX)
set(OPENCV_DATA_CONFIG_STR "${OPENCV_DATA_CONFIG_STR}
#define OPENCV_INSTALL_PREFIX \"${CMAKE_INSTALL_PREFIX}\"
")
endif()
if(OPENCV_OTHER_INSTALL_PATH)
set(OPENCV_DATA_CONFIG_STR "${OPENCV_DATA_CONFIG_STR}
#define OPENCV_DATA_INSTALL_PATH \"${OPENCV_OTHER_INSTALL_PATH}\"
")
endif()
set(OPENCV_DATA_CONFIG_STR "${OPENCV_DATA_CONFIG_STR}
#define OPENCV_BUILD_DIR \"${CMAKE_BINARY_DIR}\"
")
file(RELATIVE_PATH SOURCE_DIR_RELATIVE ${CMAKE_BINARY_DIR} ${CMAKE_SOURCE_DIR})
set(OPENCV_DATA_CONFIG_STR "${OPENCV_DATA_CONFIG_STR}
#define OPENCV_DATA_BUILD_DIR_SEARCH_PATHS \\
\"${SOURCE_DIR_RELATIVE}/\"
")
if(WIN32)
file(RELATIVE_PATH INSTALL_DATA_DIR_RELATIVE "${CMAKE_INSTALL_PREFIX}/${OPENCV_BIN_INSTALL_PATH}" "${CMAKE_INSTALL_PREFIX}/${OPENCV_OTHER_INSTALL_PATH}")
else()
file(RELATIVE_PATH INSTALL_DATA_DIR_RELATIVE "${CMAKE_INSTALL_PREFIX}/${OPENCV_LIB_INSTALL_PATH}" "${CMAKE_INSTALL_PREFIX}/${OPENCV_OTHER_INSTALL_PATH}")
endif()
list(APPEND OPENCV_INSTALL_DATA_DIR_RELATIVE "${INSTALL_DATA_DIR_RELATIVE}")
string(REPLACE ";" "\",\\\n \"" OPENCV_INSTALL_DATA_DIR_RELATIVE_STR "\"${OPENCV_INSTALL_DATA_DIR_RELATIVE}\"")
set(OPENCV_DATA_CONFIG_STR "${OPENCV_DATA_CONFIG_STR}
#define OPENCV_INSTALL_DATA_DIR_RELATIVE ${OPENCV_INSTALL_DATA_DIR_RELATIVE_STR}
")
if(EXISTS "${OPENCV_DATA_CONFIG_FILE}")
file(READ "${OPENCV_DATA_CONFIG_FILE}" __content)
endif()
if(NOT OPENCV_DATA_CONFIG_STR STREQUAL "${__content}")
file(WRITE "${OPENCV_DATA_CONFIG_FILE}" "${OPENCV_DATA_CONFIG_STR}")
endif()