Opencv-cuda error after updating cuda from version 12.3 to 12.4

OmieSawie · March 9, 2024, 10:18pm

I am getting an error when I try to run “make”

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
CMake Error at /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:230 (message):
  Could NOT find CUDA: Found unsuitable version "12.4", but required is exact
  version "12.3" (found /opt/cuda)
Call Stack (most recent call first):
  /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:598 (_FPHSA_FAILURE_MESSAGE)
  /usr/share/cmake/Modules/FindCUDA.cmake:1291 (find_package_handle_standard_args)
  /usr/lib/cmake/opencv4/OpenCVConfig.cmake:86 (find_package)
  /usr/lib/cmake/opencv4/OpenCVConfig.cmake:108 (find_host_package)
  CMakeLists.txt:3 (find_package)


-- Configuring incomplete, errors occurred!

I believe this is beacuse of updating cuda from 12.3 to 12.4.

I am on Arch and I used Pacman to install opencv-cuda and cuda, so I cannot go back to a previous version as well.

cudawarped · March 10, 2024, 5:38am

I’m not sure exactly what’s causing your error, it looks like a CMake configuration error. You probably need to clean your build directory. That said it doesn’t matter as OpenCV does not support 12.4 yet.

github.com/opencv/opencv_contrib

CUDA Toolkit 12.4.0 `tuple` incompatibility

opened 06:41PM - 08 Mar 24 UTC

runer112

bug category: build/install category: cuda

##### System information (version)  - OpenCV => 4.9.0 - Operating System / Platform => Windows 64 Bit - Compiler => Visual Studio 2022 ##### Detailed description opencv with CUDA support cannot be built using CUDA Toolkit 12.4.0. While CUDA Toolkit 12.3.2 uses thrust version 2.2.0 (https://docs.nvidia.com/cuda/archive/12.3.2/cuda-toolkit-release-notes/index.html), CUDA Toolkit 12.4.0 updates to thrust version 2.3.1 (https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html). In thrust version 2.3.0, the tuple implementation was replaced with a standard tuple implementaton (https://github.com/NVIDIA/cccl/pull/262). Notably, this changes the definition from a 10-parameter template to a variable-parameter template. So instead of a tuple of _n_ items being padded out with _10 - n_ null types to always have 10 template parameters, it now only has _n_ template parameters. This makes the function templates in cudev specified with 10 template parameters per tuple no longer viable for tuples not of size 10. An example of one such function template that's no longer viable, `cv::cudev::blockReduce`: https://github.com/opencv/opencv_contrib/blob/6b5142ff657ca676ab35233556b49a532e75e2b7/modules/cudev/include/opencv2/cudev/block/reduce.hpp#L68-L81 An example of an error I encounter: ``` [build] Z:\dev\1\opencv_contrib\modules\cudev\include\opencv2\cudev\grid\detail/reduce.hpp(379): error : no instance of overloaded function "cv::cudev::blockReduce" matches the argument list [Z:\dev\1\opencv\out\build\user\modules\world\opencv_world.vcxproj] [build] argument types are: (cuda::std::__4::tuple<volatile int *, volatile int *>, cuda::std::__4::tuple<int &, int &>, int, cuda::std::__4::tuple<cv::cudev::minimum<int>, cv::cudev::maximum<int>>) [build] blockReduce<BLOCK_SIZE>(smem_tuple(sminval, smaxval), tie(mymin, mymax), tid, make_tuple(minOp, maxOp)); [build] ^ [build] Z:\dev\1\opencv_contrib\modules\cudev\include\opencv2\cudev/block/reduce.hpp(72): note #3327-D: candidate function template "cv::cudev::blockReduce<N,P0,P1,P2,P3,P4,P5,P6,P7,P8,P9,R0,R1,R2,R3,R4,R5,R6,R7,R8,R9,Op0,Op1,Op2,Op3,Op4,Op5,Op6,Op7,Op8,Op9>(const thrust::THRUST_200301_500_520_600_610_700_750_800_860_890_900_NS::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> &, const thrust::THRUST_200301_500_520_600_610_700_750_800_860_890_900_NS::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9> &, uint, const thrust::THRUST_200301_500_520_600_610_700_750_800_860_890_900_NS::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9> &)" failed deduction [build] __declspec(__device__) __forceinline void blockReduce(const tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem, [build] ^ [build] Z:\dev\1\opencv_contrib\modules\cudev\include\opencv2\cudev/block/reduce.hpp(63): note #3327-D: candidate function template "cv::cudev::blockReduce<N,T,Op>(volatile T *, T &, uint, const Op &)" failed deduction [build] __declspec(__device__) __forceinline void blockReduce(volatile T* smem, T& val, uint tid, const Op& op) [build] ^ [build] detected during: [build] instantiation of "void cv::cudev::grid_reduce_detail::MinMaxReductor<cv::cudev::grid_reduce_detail::both, src_type, work_type>::reduceGrid<BLOCK_SIZE>(work_type *, int) [with src_type=uchar, work_type=int, BLOCK_SIZE=256]" at line 412 [build] instantiation of "void cv::cudev::grid_reduce_detail::reduce<Reductor,BLOCK_SIZE,PATCH_X,PATCH_Y,SrcPtr,ResType,MaskPtr>(SrcPtr, ResType *, MaskPtr, int, int) [with Reductor=cv::cudev::grid_reduce_detail::MinMaxReductor<cv::cudev::grid_reduce_detail::both, uchar, int>, BLOCK_SIZE=256, PATCH_X=4, PATCH_Y=4, SrcPtr=cv::cudev::GlobPtr<uchar>, ResType=int, MaskPtr=cv::cudev::WithOutMask]" at line 421 [build] instantiation of "void cv::cudev::grid_reduce_detail::reduce<Reductor,Policy,SrcPtr,ResType,MaskPtr>(const SrcPtr &, ResType *, const MaskPtr &, int, int, cudaStream_t) [with Reductor=cv::cudev::grid_reduce_detail::MinMaxReductor<cv::cudev::grid_reduce_detail::both, uchar, int>, Policy=cv::cudev::DefaultGlobReducePolicy, SrcPtr=cv::cudev::GlobPtr<uchar>, ResType=int, MaskPtr=cv::cudev::WithOutMask]" at line 460 [build] instantiation of "void cv::cudev::grid_reduce_detail::minMaxVal<Policy,SrcPtr,ResType,MaskPtr>(const SrcPtr &, ResType *, const MaskPtr &, int, int, cudaStream_t) [with Policy=cv::cudev::DefaultGlobReducePolicy, SrcPtr=cv::cudev::GlobPtr<uchar>, ResType=int, MaskPtr=cv::cudev::WithOutMask]" at line 206 of Z:\dev\1\opencv_contrib\modules\cudev\include\opencv2\cudev/grid/reduce.hpp [build] instantiation of "void cv::cudev::gridFindMinMaxVal_<Policy,SrcPtr,ResType>(const SrcPtr &, cv::cudev::GpuMat_<ResType> &, cv::cuda::Stream &) [with Policy=cv::cudev::DefaultGlobReducePolicy, SrcPtr=cv::cudev::GpuMat_<uchar>, ResType=int]" at line 349 of Z:\dev\1\opencv_contrib\modules\cudev\include\opencv2\cudev/grid/reduce.hpp [build] instantiation of "void cv::cudev::gridFindMinMaxVal(const SrcPtr &, cv::cudev::GpuMat_<ResType> &, cv::cuda::Stream &) [with SrcPtr=cv::cudev::GpuMat_<uchar>, ResType=int]" at line 68 of Z:\dev\1\opencv_contrib\modules\cudaarithm\src\cuda\minmax.cu [build] instantiation of "void <unnamed>::minMaxImpl<T,R>(const cv::cuda::GpuMat &, const cv::cuda::GpuMat &, cv::cuda::GpuMat &, cv::cuda::Stream &) [with T=uchar, R=int]" at line 92 of Z:\dev\1\opencv_contrib\modules\cudaarithm\src\cuda\minmax.cu ``` The first candidate but nonviable function template shown in the error message is the one linked above, which was viable and selected in previous CUDA Toolkit versions. I think that all templates specifying 10 template parameters per tuple can be updated to work with the new tuple definition by replacing each set of 10 template parameters with a parameter pack. I think this should still be compatible with the old tuple definition, as well. For example, I think this would be a viable implementation of `cv::cudev::blockReduce`: ```cpp template <int N, typename... P, typename... R, class... Op> __device__ __forceinline__ void blockReduce(const tuple<P...>& smem, const tuple<R...>& val, uint tid, const tuple<Op...>& op) { block_reduce_detail::Dispatcher<N>::reductor::template reduce< const tuple<P...>&, const tuple<R...>&, const tuple<Op...>&>(smem, val, tid, op); } ``` ##### Steps to reproduce Attempt to build cudev using CUDA Toolkit 12.4.0. I suspect that this error will be observed with any existing OpenCV version, OS, platform, and compiler. ##### Issue submission checklist - [x] I report the issue, it's not a question  - [x] I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution  - [x] I updated to the latest OpenCV version and the issue is still there  - [x] There is reproducer code and related data files: videos, images, onnx, etc

Topic		Replies	Views
Opencv build process keeps finding wrong version of CUDA C++ build , cuda	1	380	April 5, 2023
Use OpenCV with CUDA support in production C++ build , cuda	2	386	April 20, 2021
Compiling opencv c++ with cuda C++ windows , build , cuda	1	979	March 10, 2023
Linked opencv with CUDA supported version for a project, but it throws "throw_no_cuda" C++ build , cuda	5	2510	July 21, 2023
OpenCV C++ CUDA build error build , cuda	7	4709	May 11, 2021

Opencv-cuda error after updating cuda from version 12.3 to 12.4

Related Topics