System Information

OpenCV version: 4.10.0, 4.x (branch)
Operating System: Gentoo Linux
CMake: 3.30.0
GPU: GeForce GTX 960M
Native CUDA arch: 50

Detailed description

When configuring OpenCV in the clean build directory, if the ENABLE_CUDA_FIRST_CLASS_LANGUAGE option is set to yes and the MIN_VER_CMAKE is set to 3.18 or higher, CMake ignores values of CUDA_ARCH_BIN and CUDA_ARCH_PTX options.

In this case I only can set NVIDIA GPU arch (OPENCV_CUDA_ARCH_BIN) and NVIDIA PTX archs (OPENCV_CUDA_ARCH_PTX) CMake variables using the CUDAARCHS environment variable.

Steps to reproduce

Note: to reveal the issue on your system, please replace 5.0 with any value different from what OpenCV detects out of the box.

tar -xpf opencv-4.10.0.tar.gz
tar -xpf opencv_contrib-4.10.0.tar.gz
mkdir build
cd build
cmake -DMIN_VER_CMAKE=3.18 -DWITH_CUDA=yes -DCUDA_ARCH_BIN=5.0 -DCUDA_ARCH_PTX=5.0 -DENABLE_CUDA_FIRST_CLASS_LANGUAGE=yes -DOPENCV_EXTRA_MODULES_PATH=.
./opencv_contrib-4.10.0/modules ../opencv-4.10.0 | grep -A2 'NVIDIA CUDA'

In my case, it returns this:

CMake Warning (dev) at CMakeLists.txt:127 (enable_language):
  project() should be called prior to this enable_language() call.
This warning is for project developers.  Use -Wno-dev to suppress it.

CMake Warning at /dev/shm/opencv_contrib-4.10.0/modules/cudacodec/CMakeLists.txt:26 (message):
  cudacodec::VideoReader requires Nvidia Video Codec SDK.  Please resolve
  dependency or disable WITH_NVCUVID=OFF


CMake Warning at /dev/shm/opencv_contrib-4.10.0/modules/cudacodec/CMakeLists.txt:30 (message):
  cudacodec::VideoWriter requires Nvidia Video Codec SDK.  Please resolve
  dependency or disable WITH_NVCUVENC=OFF

--   NVIDIA CUDA:                   YES (ver 12.5.40, CUFFT CUBLAS)
--     NVIDIA GPU arch:             52
--     NVIDIA PTX archs:            52

The same happens with the latest snapshots of 4.x branches of opencv and opencv_contrib repositories.

In case I remove one or both of MIN_VER_CMAKE and ENABLE_CUDA_FIRST_CLASS_LANGUAGE, it works as expected (with the same warnings).

$ cmake -DWITH_CUDA=yes -DCUDA_ARCH_BIN=5.0 -DCUDA_ARCH_PTX=5.0 -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.10.0/modules ../opencv-4.10.0 2>/dev/null | grep -A2 'NVIDIA CUDA'
--   NVIDIA CUDA:                   YES (ver 12.5, CUFFT CUBLAS)
--     NVIDIA GPU arch:             50
--     NVIDIA PTX archs:            50

The CUDAARCHS environment variable also helps.

$ env CUDAARCHS=50 cmake -DMIN_VER_CMAKE=3.18 -DWITH_CUDA=yes -DCUDA_ARCH_BIN=5.0 -DCUDA_ARCH_PTX=5.0 -DENABLE_CUDA_FIRST_CLASS_LANGUAGE=yes -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.10.0/modules ../opencv-4.10.0 2>/dev/null | grep -A2 'NVIDIA CUDA'
--   NVIDIA CUDA:                   YES (ver 12.5.40, CUFFT CUBLAS)
--     NVIDIA GPU arch:             50
--     NVIDIA PTX archs:            50

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
• I updated to the latest OpenCV version and the issue is still there
• There is reproducer code and related data files (videos, images, onnx, etc)