mirror of
https://github.com/opencv/opencv.git
synced 2026-07-31 00:03:03 +04:00
Merge pull request #328 from jet47:new-gpu-fixes
This commit is contained in:
cuda-geek
OpenCV Buildbot
@@ -3,17 +3,17 @@ if(${CMAKE_VERSION} VERSION_LESS "2.8.3")
|
||||
return()
|
||||
endif()
|
||||
|
||||
if (WIN32 AND NOT MSVC)
|
||||
message(STATUS "CUDA compilation is disabled (due to only Visual Studio compiler suppoted on your platform).")
|
||||
if(WIN32 AND NOT MSVC)
|
||||
message(STATUS "CUDA compilation is disabled (due to only Visual Studio compiler supported on your platform).")
|
||||
return()
|
||||
endif()
|
||||
|
||||
if (CMAKE_COMPILER_IS_GNUCXX AND NOT APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
|
||||
message(STATUS "CUDA compilation is disabled (due to Clang unsuppoted on your platform).")
|
||||
if(CMAKE_COMPILER_IS_GNUCXX AND NOT APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
|
||||
message(STATUS "CUDA compilation is disabled (due to Clang unsupported on your platform).")
|
||||
return()
|
||||
endif()
|
||||
|
||||
find_package(CUDA 4.1)
|
||||
find_package(CUDA 4.2 QUIET)
|
||||
|
||||
if(CUDA_FOUND)
|
||||
set(HAVE_CUDA 1)
|
||||
@@ -26,15 +26,20 @@ if(CUDA_FOUND)
|
||||
set(HAVE_CUBLAS 1)
|
||||
endif()
|
||||
|
||||
message(STATUS "CUDA detected: " ${CUDA_VERSION})
|
||||
|
||||
if(${CUDA_VERSION_STRING} VERSION_GREATER "4.1")
|
||||
set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
|
||||
else()
|
||||
set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0)" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
|
||||
if(WITH_NVCUVID)
|
||||
find_cuda_helper_libs(nvcuvid)
|
||||
set(HAVE_NVCUVID 1)
|
||||
endif()
|
||||
|
||||
set(CUDA_ARCH_PTX "2.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
|
||||
message(STATUS "CUDA detected: " ${CUDA_VERSION})
|
||||
|
||||
if (CARMA)
|
||||
set(CUDA_ARCH_BIN "2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
|
||||
set(CUDA_ARCH_PTX "3.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
|
||||
else()
|
||||
set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
|
||||
set(CUDA_ARCH_PTX "2.0 3.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
|
||||
endif()
|
||||
|
||||
string(REGEX REPLACE "\\." "" ARCH_BIN_NO_POINTS "${CUDA_ARCH_BIN}")
|
||||
string(REGEX REPLACE "\\." "" ARCH_PTX_NO_POINTS "${CUDA_ARCH_PTX}")
|
||||
@@ -72,11 +77,20 @@ if(CUDA_FOUND)
|
||||
|
||||
# Tell NVCC to add PTX intermediate code for the specified architectures
|
||||
string(REGEX MATCHALL "[0-9]+" ARCH_LIST "${ARCH_PTX_NO_POINTS}")
|
||||
foreach(ARCH IN LISTS ARCH_LIST)
|
||||
set(NVCC_FLAGS_EXTRA ${NVCC_FLAGS_EXTRA} -gencode arch=compute_${ARCH},code=compute_${ARCH})
|
||||
set(OPENCV_CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX} ${ARCH}")
|
||||
set(OPENCV_CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES} ${ARCH}")
|
||||
endforeach()
|
||||
foreach(ARCH IN LISTS ARCH_LIST)
|
||||
set(NVCC_FLAGS_EXTRA ${NVCC_FLAGS_EXTRA} -gencode arch=compute_${ARCH},code=compute_${ARCH})
|
||||
set(OPENCV_CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX} ${ARCH}")
|
||||
set(OPENCV_CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES} ${ARCH}")
|
||||
endforeach()
|
||||
|
||||
if(CARMA)
|
||||
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} --target-cpu-architecture=ARM" )
|
||||
|
||||
if (CMAKE_VERSION VERSION_LESS 2.8.10)
|
||||
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -ccbin=${CMAKE_CXX_COMPILER}" )
|
||||
endif()
|
||||
|
||||
endif()
|
||||
|
||||
# These vars will be processed in other scripts
|
||||
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ${NVCC_FLAGS_EXTRA})
|
||||
@@ -84,7 +98,7 @@ if(CUDA_FOUND)
|
||||
|
||||
message(STATUS "CUDA NVCC target flags: ${CUDA_NVCC_FLAGS}")
|
||||
|
||||
OCV_OPTION(CUDA_FAST_MATH "Enable --use_fast_math for CUDA compiler " OFF)
|
||||
OCV_OPTION(CUDA_FAST_MATH "Enable --use_fast_math for CUDA compiler " OFF)
|
||||
|
||||
if(CUDA_FAST_MATH)
|
||||
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --use_fast_math)
|
||||
@@ -92,7 +106,6 @@ if(CUDA_FOUND)
|
||||
|
||||
mark_as_advanced(CUDA_BUILD_CUBIN CUDA_BUILD_EMULATION CUDA_VERBOSE_BUILD CUDA_SDK_ROOT_DIR)
|
||||
|
||||
unset(CUDA_npp_LIBRARY CACHE)
|
||||
find_cuda_helper_libs(npp)
|
||||
|
||||
macro(ocv_cuda_compile VAR)
|
||||
@@ -106,15 +119,15 @@ if(CUDA_FOUND)
|
||||
string(REPLACE "-ggdb3" "" ${var} "${${var}}")
|
||||
endforeach()
|
||||
|
||||
if (BUILD_SHARED_LIBS)
|
||||
if(BUILD_SHARED_LIBS)
|
||||
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -DCVAPI_EXPORTS)
|
||||
endif()
|
||||
|
||||
if(UNIX OR APPLE)
|
||||
set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fPIC)
|
||||
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fPIC)
|
||||
endif()
|
||||
if(APPLE)
|
||||
set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fno-finite-math-only)
|
||||
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fno-finite-math-only)
|
||||
endif()
|
||||
|
||||
# disabled because of multiple warnings during building nvcc auto generated files
|
||||
|
||||
@@ -42,8 +42,9 @@
|
||||
set(OpenCV_COMPUTE_CAPABILITIES @OpenCV_CUDA_CC_CONFIGCMAKE@)
|
||||
|
||||
set(OpenCV_CUDA_VERSION @OpenCV_CUDA_VERSION@)
|
||||
set(OpenCV_USE_CUBLAS @HAVE_CUBLAS@)
|
||||
set(OpenCV_USE_CUFFT @HAVE_CUFFT@)
|
||||
set(OpenCV_USE_CUBLAS @HAVE_CUBLAS@)
|
||||
set(OpenCV_USE_CUFFT @HAVE_CUFFT@)
|
||||
set(OpenCV_USE_NVCUVID @HAVE_NVCUVID@)
|
||||
|
||||
# Android API level from which OpenCV has been compiled is remembered
|
||||
set(OpenCV_ANDROID_NATIVE_API_LEVEL @OpenCV_ANDROID_NATIVE_API_LEVEL_CONFIGCMAKE@)
|
||||
@@ -218,17 +219,22 @@ foreach(__opttype OPT DBG)
|
||||
else()
|
||||
#TODO: duplicates are annoying but they should not be the problem
|
||||
endif()
|
||||
# fix hard coded paths for CUDA libraries under Windows
|
||||
if(WIN32 AND OpenCV_CUDA_VERSION AND NOT OpenCV_SHARED)
|
||||
|
||||
# CUDA
|
||||
if(OpenCV_CUDA_VERSION AND (CARMA OR (WIN32 AND NOT OpenCV_SHARED)))
|
||||
if(NOT CUDA_FOUND)
|
||||
find_package(CUDA ${OpenCV_CUDA_VERSION} EXACT REQUIRED)
|
||||
else()
|
||||
if(NOT CUDA_VERSION_STRING VERSION_EQUAL OpenCV_CUDA_VERSION)
|
||||
message(FATAL_ERROR "OpenCV static library compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
|
||||
if(WIN32)
|
||||
message(FATAL_ERROR "OpenCV static library was compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
|
||||
else()
|
||||
message(FATAL_ERROR "OpenCV library for CARMA was compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
|
||||
endif()
|
||||
endif()
|
||||
endif()
|
||||
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY} ${CUDA_nvcuvid_LIBRARY} ${CUDA_nvcuvenc_LIBRARY})
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
|
||||
|
||||
if(OpenCV_USE_CUBLAS)
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_CUBLAS_LIBRARIES})
|
||||
@@ -238,6 +244,13 @@ foreach(__opttype OPT DBG)
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_CUFFT_LIBRARIES})
|
||||
endif()
|
||||
|
||||
if(OpenCV_USE_NVCUVID)
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_nvcuvid_LIBRARIES})
|
||||
endif()
|
||||
|
||||
if(WIN32)
|
||||
list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_nvcuvenc_LIBRARIES})
|
||||
endif()
|
||||
endif()
|
||||
endforeach()
|
||||
|
||||
|
||||
@@ -175,21 +175,15 @@
|
||||
/* NVidia Cuda Runtime API*/
|
||||
#cmakedefine HAVE_CUDA
|
||||
|
||||
/* OpenCL Support */
|
||||
#cmakedefine HAVE_OPENCL
|
||||
|
||||
/* AMD's OpenCL Fast Fourier Transform Library*/
|
||||
#cmakedefine HAVE_CLAMDFFT
|
||||
|
||||
/* AMD's Basic Linear Algebra Subprograms Library*/
|
||||
#cmakedefine HAVE_CLAMDBLAS
|
||||
|
||||
/* NVidia Cuda Fast Fourier Transform (FFT) API*/
|
||||
#cmakedefine HAVE_CUFFT
|
||||
|
||||
/* NVidia Cuda Basic Linear Algebra Subprograms (BLAS) API*/
|
||||
#cmakedefine HAVE_CUBLAS
|
||||
|
||||
/* NVidia Video Decoding API*/
|
||||
#cmakedefine HAVE_NVCUVID
|
||||
|
||||
/* Compile for 'real' NVIDIA GPU architectures */
|
||||
#define CUDA_ARCH_BIN "${OPENCV_CUDA_ARCH_BIN}"
|
||||
|
||||
@@ -202,6 +196,15 @@
|
||||
/* Create PTX or BIN for 1.0 compute capability */
|
||||
#cmakedefine CUDA_ARCH_BIN_OR_PTX_10
|
||||
|
||||
/* OpenCL Support */
|
||||
#cmakedefine HAVE_OPENCL
|
||||
|
||||
/* AMD's OpenCL Fast Fourier Transform Library*/
|
||||
#cmakedefine HAVE_CLAMDFFT
|
||||
|
||||
/* AMD's Basic Linear Algebra Subprograms Library*/
|
||||
#cmakedefine HAVE_CLAMDBLAS
|
||||
|
||||
/* VideoInput library */
|
||||
#cmakedefine HAVE_VIDEOINPUT
|
||||
|
||||
|
||||
Reference in New Issue
Block a user