Add support for compute_10

[jramapuram]

-DCMAKE_CUDA_ARCHITECTURES=CUDA_ARCHITECTURES=80;90;10 fails

+ cmake -B build . -DFAISS_ENABLE_GPU=ON -DFAISS_ENABLE_PYTHON=ON -DFAISS_ENABLE_RAFT=OFF -DBUILD_TESTING=OFF -DBUILD_SHARED_LIBS=ON -DFAISS_ENABLE_C_API=ON -DCMAKE_BUILD_TYPE=Release -DFAISS_OPT_LEVEL=avx2 -DCUDAToolkit_ROOT=/usr/local/cuda -DBLA_VENDOR=Intel10_64_dyn -DMKL_LIBRARIES=/miniconda/lib '-DCMAKE_CUDA_ARCHITECTURES=80;90;10'
-- The CXX compiler identification is GNU 13.3.0
-- The CUDA compiler identification is NVIDIA 12.9.86 with host compiler GNU 13.3.0
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - failed
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc - broken
CMake Error at /miniconda/lib/python3.11/site-packages/cmake/data/share/cmake-4.1/Modules/CMakeTestCUDACompiler.cmake:59 (message):
  The CUDA compiler

    "/usr/local/cuda/bin/nvcc"

  is not able to compile a simple test program.

  It fails with the following output:

    Change Dir: '/tmp/tmp.VYCSV6h5jb/faiss/build/CMakeFiles/CMakeScratch/TryCompile-Lb1FVu'
    
    Run Build Command(s): /miniconda/lib/python3.11/site-packages/cmake/data/bin/cmake -E env VERBOSE=1 /usr/bin/gmake -f Makefile cmTC_2a2ac/fast
    /usr/bin/gmake  -f CMakeFiles/cmTC_2a2ac.dir/build.make CMakeFiles/cmTC_2a2ac.dir/build
    gmake[1]: Entering directory '/tmp/tmp.VYCSV6h5jb/faiss/build/CMakeFiles/CMakeScratch/TryCompile-Lb1FVu'
    Building CUDA object CMakeFiles/cmTC_2a2ac.dir/main.cu.o
    /usr/local/cuda/bin/nvcc -forward-unknown-to-host-compiler   "--generate-code=arch=compute_80,code=[compute_80,sm_80]" "--generate-code=arch=compute_90,code=[compute_90,sm_90]" "--generate-code=arch=compute_10,code=[compute_10,sm_10]" -MD -MT CMakeFiles/cmTC_2a2ac.dir/main.cu.o -MF CMakeFiles/cmTC_2a2ac.dir/main.cu.o.d -x cu -c /tmp/tmp.VYCSV6h5jb/faiss/build/CMakeFiles/CMakeScratch/TryCompile-Lb1FVu/main.cu -o CMakeFiles/cmTC_2a2ac.dir/main.cu.o
    nvcc fatal   : Unsupported gpu architecture 'compute_10'
    gmake[1]: *** [CMakeFiles/cmTC_2a2ac.dir/build.make:82: CMakeFiles/cmTC_2a2ac.dir/main.cu.o] Error 1
    gmake[1]: Leaving directory '/tmp/tmp.VYCSV6h5jb/faiss/build/CMakeFiles/CMakeScratch/TryCompile-Lb1FVu'
    gmake: *** [Makefile:134: cmTC_2a2ac/fast] Error 2
    
    



  CMake will not be able to correctly generate this project.
Call Stack (most recent call first):
  CMakeLists.txt:49 (project)

[trang-nm-nguyen]

We don't have plan to support this feature yet, but feel free to make contribution and we will help with code review!

[mnorris11]

Is the CUDA_ARCHITECTURES value for Blackwell supposed to be 10 or 100? I actually didn't find good documentation on it with an admittedly brief search. But regardless, like Trang said, we don't immediately plan to update, but will happily accept contributions. The pinned nvcc version probably would need to get updated.