Request: Support for build cuda support for libvmaf using ffnvcodec

[ghost]

This will simplify native compilation and make cross-compilation possible.

The patch for ffmpeg vf_libvmaf_cuda seems to have been sent to ffmpeg-devel, and I don't think anyone wants to use both the cuda sdk and the ffnvcodec in a ffmpeg build.

[kylophone]

Yes, I think this is probably the best way forward, I was planning on getting to this eventually. If you (or anyone else) wants to work on it in the meantime, please let me know.

[gedoensmax]

So you need mvcc to build the CUDA code but it does not require CUDA runtime and linking against CUDA driver stubs that are for example included in a CUDA docker container works well.