Administrator/opencv-MIRROR
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-09-11 04:43:22 -05:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #29658 from SavyaSanchi-Sharma:cudnnjit

Browse Source 
This PR is about Introducing cuDNN JIT support for the DNN Module

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
This commit is contained in:
Savya Sanchi Sharma
2026-08-27 12:01:21 +05:30
committed by GitHub
parent 71a601ea0e
commit e16382025c
9 changed files with 136 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
CMakeLists.txt
View File

@@ -204,6 +204,9 @@ OCV_OPTION(WITH_CUBLAS "Include NVidia Cuda Basic Linear Algebra Subprograms (BL
OCV_OPTION(WITH_CUDNN "Include NVIDIA CUDA Deep Neural Network (cuDNN) library support" WITH_CUDA
VISIBLE_IF WITH_CUDA
VERIFY HAVE_CUDNN)
OCV_OPTION(WITH_CUDNNJIT "Include NVIDIA CUDA Deep Neural Network (cuDNN) graph API support with runtime compiled (JIT) engines, used only when WITH_CUDNN is off" OFF
VISIBLE_IF WITH_CUDA
VERIFY HAVE_CUDNNJIT OR HAVE_CUDNN)
OCV_OPTION(WITH_NVCUVID "Include NVidia Video Decoding library support" ON
VISIBLE_IF WITH_CUDA
VERIFY HAVE_NVCUVID)
@@ -1912,6 +1915,11 @@ if(WITH_CUDA OR HAVE_CUDA)
status(" cuDNN:" HAVE_CUDNN THEN "YES (ver ${CUDNN_VERSION})" ELSE NO)
endif()
if(WITH_CUDNNJIT OR HAVE_CUDNNJIT)
status("")
status(" cuDNN JIT:" HAVE_CUDNNJIT THEN "YES (ver ${CUDNN_VERSION})" ELSE NO)
endif()
if(WITH_VULKAN OR HAVE_VULKAN)
status("")
status(" Vulkan:" HAVE_VULKAN THEN "YES" ELSE "NO")

71
cmake/FindCUDNNJIT.cmake Normal file
View File

@@ -0,0 +1,71 @@
# template taken from https://cmake.org/cmake/help/v3.14/manual/cmake-developer.7.html
# Distributed under the OSI-approved BSD 3-Clause License. See accompanying
# file Copyright.txt or https://cmake.org/licensing for details.
if(CUDA_FOUND)
find_cuda_helper_libs(cudnn)
find_cuda_helper_libs(cudnn_graph)
find_cuda_helper_libs(cudnn_engines_runtime_compiled)
set(CUDNN_SHIM_LIBRARY ${CUDA_cudnn_LIBRARY} CACHE FILEPATH "location of the cuDNN dispatch shim library")
set(CUDNN_GRAPH_LIBRARY ${CUDA_cudnn_graph_LIBRARY} CACHE FILEPATH "location of the cuDNN graph library")
set(CUDNN_ENGINES_RTC_LIBRARY ${CUDA_cudnn_engines_runtime_compiled_LIBRARY} CACHE FILEPATH "location of the cuDNN runtime compiled engines library")
unset(CUDA_cudnn_LIBRARY CACHE)
unset(CUDA_cudnn_graph_LIBRARY CACHE)
unset(CUDA_cudnn_engines_runtime_compiled_LIBRARY CACHE)
endif()
if(CUDNN_GRAPH_LIBRARY)
find_path(CUDNNJIT_INCLUDE_DIR
cudnn_graph.h
PATHS ${CUDA_TOOLKIT_INCLUDE}
DOC "location of cudnn_graph.h"
NO_DEFAULT_PATH
)
if(NOT CUDNNJIT_INCLUDE_DIR)
find_path(CUDNNJIT_INCLUDE_DIR
cudnn_graph.h
DOC "location of cudnn_graph.h"
)
endif()
endif()
if(CUDNNJIT_INCLUDE_DIR AND EXISTS "${CUDNNJIT_INCLUDE_DIR}/cudnn_version.h")
file(READ "${CUDNNJIT_INCLUDE_DIR}/cudnn_version.h" CUDNN_H_CONTENTS)
string(REGEX MATCH "define CUDNN_MAJOR ([0-9]+)" _ "${CUDNN_H_CONTENTS}")
set(CUDNN_VERSION_MAJOR ${CMAKE_MATCH_1} CACHE INTERNAL "")
string(REGEX MATCH "define CUDNN_MINOR ([0-9]+)" _ "${CUDNN_H_CONTENTS}")
set(CUDNN_VERSION_MINOR ${CMAKE_MATCH_1} CACHE INTERNAL "")
string(REGEX MATCH "define CUDNN_PATCHLEVEL ([0-9]+)" _ "${CUDNN_H_CONTENTS}")
set(CUDNN_VERSION_PATCH ${CMAKE_MATCH_1} CACHE INTERNAL "")
set(CUDNN_VERSION "${CUDNN_VERSION_MAJOR}.${CUDNN_VERSION_MINOR}.${CUDNN_VERSION_PATCH}")
unset(CUDNN_H_CONTENTS)
endif()
include(FindPackageHandleStandardArgs)
find_package_handle_standard_args(CUDNNJIT
FOUND_VAR CUDNNJIT_FOUND
REQUIRED_VARS
CUDNN_SHIM_LIBRARY
CUDNN_GRAPH_LIBRARY
CUDNN_ENGINES_RTC_LIBRARY
CUDNNJIT_INCLUDE_DIR
VERSION_VAR CUDNN_VERSION
)
if(CUDNNJIT_FOUND)
set(CUDNNJIT_LIBRARIES ${CUDNN_SHIM_LIBRARY} ${CUDNN_GRAPH_LIBRARY} ${CUDNN_ENGINES_RTC_LIBRARY})
set(CUDNNJIT_INCLUDE_DIRS ${CUDNNJIT_INCLUDE_DIR})
endif()
mark_as_advanced(
CUDNN_SHIM_LIBRARY
CUDNN_GRAPH_LIBRARY
CUDNN_ENGINES_RTC_LIBRARY
CUDNNJIT_INCLUDE_DIR
CUDNN_VERSION
)

28
cmake/OpenCVDetectCUDA.cmake
View File

@@ -65,6 +65,20 @@ if(WITH_CUDNN)
endif()
endif()
if(WITH_CUDNNJIT)
if(HAVE_CUDNN)
message(STATUS "CUDA: cuDNN found, ignoring WITH_CUDNNJIT. Configure with -DWITH_CUDNN=OFF to use the cuDNN graph API instead.")
else()
set(CMAKE_MODULE_PATH "${OpenCV_SOURCE_DIR}/cmake" ${CMAKE_MODULE_PATH})
find_host_package(CUDNNJIT "${MIN_VER_CUDNNJIT}")
list(REMOVE_AT CMAKE_MODULE_PATH 0)
if(CUDNNJIT_FOUND)
set(HAVE_CUDNNJIT 1)
endif()
endif()
endif()
include(cmake/OpenCVDetectCUDAUtils.cmake)
if(WITH_NVCUVID OR WITH_NVCUVENC)
@@ -195,6 +209,13 @@ if(HAVE_CUDA)
endforeach()
endif()
if(HAVE_CUDNNJIT)
foreach(p ${CUDNNJIT_LIBRARIES})
get_filename_component(_tmp ${p} PATH)
list(APPEND CUDA_LIBS_PATH ${_tmp})
endforeach()
endif()
if(HAVE_CUFFT)
foreach(p ${CUDA_cufft_LIBRARY})
get_filename_component(_tmp ${p} PATH)
@@ -217,6 +238,10 @@ if(HAVE_CUDA)
set(CUDNN_LIBRARIES_ABS ${CUDNN_LIBRARIES})
ocv_convert_to_lib_name(CUDNN_LIBRARIES ${CUDNN_LIBRARIES})
endif()
if(HAVE_CUDNNJIT)
set(CUDNNJIT_LIBRARIES_ABS ${CUDNNJIT_LIBRARIES})
ocv_convert_to_lib_name(CUDNNJIT_LIBRARIES ${CUDNNJIT_LIBRARIES})
endif()
if(HAVE_CUFFT)
set(CUDA_cufft_LIBRARY_ABS ${CUDA_cufft_LIBRARY})
ocv_convert_to_lib_name(CUDA_cufft_LIBRARY ${CUDA_cufft_LIBRARY})
@@ -244,6 +269,9 @@ if(HAVE_CUDA)
if(HAVE_CUDNN)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDNN_LIBRARIES})
endif()
if(HAVE_CUDNNJIT)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDNNJIT_LIBRARIES})
endif()
if(HAVE_CUFFT)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cufft_LIBRARY})
endif()

14
cmake/OpenCVDetectCUDALanguage.cmake
View File

@@ -71,6 +71,20 @@ if(WITH_CUDNN)
endif()
endif()
if(WITH_CUDNNJIT)
if(HAVE_CUDNN)
message(STATUS "CUDA: cuDNN found, ignoring WITH_CUDNNJIT. Configure with -DWITH_CUDNN=OFF to use the cuDNN graph API instead.")
else()
set(CMAKE_MODULE_PATH "${OpenCV_SOURCE_DIR}/cmake" ${CMAKE_MODULE_PATH})
find_host_package(CUDNNJIT "${MIN_VER_CUDNNJIT}")
list(REMOVE_AT CMAKE_MODULE_PATH 0)
if(CUDNNJIT_FOUND)
set(HAVE_CUDNNJIT 1)
endif()
endif()
endif()
if(WITH_NVCUVID OR WITH_NVCUVENC)
ocv_check_for_nvidia_video_codec_sdk("${CUDAToolkit_LIBRARY_ROOT}")
endif()

1
cmake/OpenCVMinDepVersions.cmake
View File

@@ -4,5 +4,6 @@ if(NOT DEFINED MIN_VER_CMAKE)
endif()
set(MIN_VER_CUDA 6.5)
set(MIN_VER_CUDNN 7.5)
set(MIN_VER_CUDNNJIT 9.0)
set(MIN_VER_PYTHON3 3.2)
set(MIN_VER_ZLIB 1.2.3)

1
cmake/templates/cvconfig.h.in
View File

@@ -36,6 +36,7 @@
/* NVIDIA CUDA Deep Neural Network (cuDNN) API*/
#cmakedefine HAVE_CUDNN
#cmakedefine HAVE_CUDNNJIT
/* NVIDIA CUDA Fast Fourier Transform (FFT) API*/
#cmakedefine HAVE_CUFFT

14
modules/dnn/CMakeLists.txt
View File

@@ -51,19 +51,19 @@ endif()
ocv_option(OPENCV_DNN_CUDA "Build with CUDA support"
HAVE_CUDA
AND HAVE_CUBLAS
AND HAVE_CUDNN
AND (HAVE_CUDNN OR HAVE_CUDNNJIT)
)
if(OPENCV_DNN_CUDA)
if(HAVE_CUDA AND HAVE_CUBLAS AND HAVE_CUDNN)
if(HAVE_CUDA AND HAVE_CUBLAS AND (HAVE_CUDNN OR HAVE_CUDNNJIT))
ocv_target_compile_definitions(${the_module} PRIVATE "CV_CUDA4DNN=1")
else()
if(NOT HAVE_CUDA)
message(SEND_ERROR "DNN: CUDA backend requires CUDA Toolkit. Please resolve dependency or disable OPENCV_DNN_CUDA=OFF")
elseif(NOT HAVE_CUBLAS)
message(SEND_ERROR "DNN: CUDA backend requires cuBLAS. Please resolve dependency or disable OPENCV_DNN_CUDA=OFF")
elseif(NOT HAVE_CUDNN)
message(SEND_ERROR "DNN: CUDA backend requires cuDNN. Please resolve dependency or disable OPENCV_DNN_CUDA=OFF")
elseif(NOT HAVE_CUDNN AND NOT HAVE_CUDNNJIT)
message(SEND_ERROR "DNN: CUDA backend requires cuDNN (WITH_CUDNN) or the cuDNN graph API (WITH_CUDNNJIT). Please resolve dependency or disable OPENCV_DNN_CUDA=OFF")
endif()
endif()
endif()
@@ -461,8 +461,8 @@ else()
set(sources_options EXCLUDE_OPENCL)
endif()
if(OPENCV_DNN_CUDA AND HAVE_CUDA AND HAVE_CUBLAS AND HAVE_CUDNN)
list(APPEND include_dirs ${CUDA_TOOLKIT_INCLUDE} ${CUDNN_INCLUDE_DIRS})
if(OPENCV_DNN_CUDA AND HAVE_CUDA AND HAVE_CUBLAS AND (HAVE_CUDNN OR HAVE_CUDNNJIT))
list(APPEND include_dirs ${CUDA_TOOLKIT_INCLUDE} ${CUDNN_INCLUDE_DIRS} ${CUDNNJIT_INCLUDE_DIRS})
set(CC_LIST ${CUDA_ARCH_BIN})
separate_arguments(CC_LIST)
foreach(cc ${CC_LIST})
@@ -472,7 +472,7 @@ if(OPENCV_DNN_CUDA AND HAVE_CUDA AND HAVE_CUBLAS AND HAVE_CUDNN)
endforeach()
unset(CC_LIST)
if(ENABLE_CUDA_FIRST_CLASS_LANGUAGE)
list(APPEND libs CUDA::cudart${CUDA_LIB_EXT} ${CUDNN_LIBRARIES} CUDA::cublas${CUDA_LIB_EXT})
list(APPEND libs CUDA::cudart${CUDA_LIB_EXT} ${CUDNN_LIBRARIES} ${CUDNNJIT_LIBRARIES} CUDA::cublas${CUDA_LIB_EXT})
if(NOT CUDA_VERSION VERSION_LESS 10.1)
list(APPEND libs CUDA::cublasLt${CUDA_LIB_EXT})
endif()

2
modules/dnn/cmake/hooks/INIT_MODULE_SOURCES_opencv_dnn.cmake
View File

@@ -4,7 +4,7 @@ if(NOT (OPENCV_DNN_OPENCL AND HAVE_OPENCL))
ocv_list_filterout(OPENCV_MODULE_${the_module}_HEADERS "/ocl4dnn/")
endif()
if(NOT (OPENCV_DNN_CUDA AND HAVE_CUDA AND HAVE_CUBLAS AND HAVE_CUDNN))
if(NOT (OPENCV_DNN_CUDA AND HAVE_CUDA AND HAVE_CUBLAS AND (HAVE_CUDNN OR HAVE_CUDNNJIT)))
message(STATUS "opencv_dnn: filter out cuda4dnn source code")
ocv_list_filterout(OPENCV_MODULE_${the_module}_SOURCES "/cuda4dnn/")
ocv_list_filterout(OPENCV_MODULE_${the_module}_HEADERS "/cuda4dnn/")

5
modules/dnn/src/cuda4dnn/csl/cudnn/cudnn.hpp
View File

@@ -7,7 +7,12 @@
#include "../pointer.hpp"
#ifdef HAVE_CUDNN
#include <cudnn.h>
#elif defined(HAVE_CUDNNJIT)
#include <cudnn_graph.h>
#include <cudnn_ops.h>
#endif
#include <cstddef>
#include <array>