Closed skiminki closed 4 years ago
Note: the CI build failure doesn't seem to be caused by my patch. My guess is that the CI server has updated CUDA version for Windows and now the NVCC path does not match anymore with the path hardcoded in lib/cuda.pri. My patch shouldn't be changing anything for the Windows paths.
The failure is: NMAKE : fatal error U1077: '"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin\nvcc.EXE"' : return code '0x1'
The path looks as intended.
Add support for probing the CUDA directory locations for include and lib64, and bail out if CUDA was not found. Use /usr/local/cuda and /opt/cuda as the CUDA bases for probing.
Make nvcc overridable.
Add a quick start guide for building and running on Linux in README.md