Skip to content

[Build] NVidia RTX 6000 Pro / CUDA 13.1 / CUDA_architecture 12.0 not supported #3695

@ktdq

Description

@ktdq

CUDA 13.1 is required to use RTX 6000 Pro on GCP, but ArrayFire build fails, with probably more errors down the line:

/home/j/arrayfire/arrayfire/src/backend/cuda/cufft.cu(41): error: identifier "CUFFT_INCOMPLETE_PARAMETER_LIST" is undefined
          case CUFFT_INCOMPLETE_PARAMETER_LIST:
               ^

/home/j/arrayfire/arrayfire/src/backend/cuda/cufft.cu(47): error: identifier "CUFFT_PARSE_ERROR" is undefined
          case CUFFT_PARSE_ERROR: return "cuFFT: plan parse error";
               ^

/home/j/arrayfire/arrayfire/src/backend/cuda/cufft.cu(53): error: identifier "CUFFT_LICENSE_ERROR" is undefined
          case CUFFT_LICENSE_ERROR: return "cuFFT: license error";
               ^

3 errors detected in the compilation of "/home/j/arrayfire/arrayfire/src/backend/cuda/cufft.cu".
CMake Error at /home/j/arrayfire2/lib/python3.9/site-packages/cmake/data/share/cmake-4.2/Modules/FindCUDA/select_compute_arch.cmake:245 (message):
  Unknown CUDA Architecture Name 12.0 in CUDA_SELECT_NVCC_ARCH_FLAGS
Call Stack (most recent call first):
  CMakeModules/AFcuda_helpers.cmake:9 (cuda_select_nvcc_arch_flags)
  src/backend/cuda/CMakeLists.txt:18 (include)

CMake Error at /home/j/arrayfire2/lib/python3.9/site-packages/cmake/data/share/cmake-4.2/Modules/FindCUDA/select_compute_arch.cmake:249 (message):
  arch_bin wasn't set for some reason
Call Stack (most recent call first):
  CMakeModules/AFcuda_helpers.cmake:9 (cuda_select_nvcc_arch_flags)
  src/backend/cuda/CMakeLists.txt:18 (include)

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions