Open gmarkall opened 2 months ago
See: https://github.com/NVIDIA/numba-cuda/blob/e3658d4ac6475d8d7f615aefb2e07e39aee3912e/numba_cuda/numba/cuda/codegen.py#L27-L29
It would be better to search for CUDA command line tools in the more normal way, e.g. from CUDA_HOME / CUDA_PATH first.
CUDA_HOME
CUDA_PATH
It'd probably make sense to add logic for the search in cuda_paths.py alongside the nvvm and libdevice searches.
cuda_paths.py
https://github.com/NVIDIA/numba-cuda/blob/main/numba_cuda/numba/cuda/cuda_paths.py
See: https://github.com/NVIDIA/numba-cuda/blob/e3658d4ac6475d8d7f615aefb2e07e39aee3912e/numba_cuda/numba/cuda/codegen.py#L27-L29
It would be better to search for CUDA command line tools in the more normal way, e.g. from
CUDA_HOME
/CUDA_PATH
first.