Nevertheless, still worth mentioning a 2017 change about ffmpeg and cuda.
So, noting a reference to building ffmpeg code with cuda/cuvid natively/easily and with opencl with a bit of fiddling
https://ffmpeg.zeranoe.com/forum/viewto ... 073#p12073
... DGDecNV etc are the gold standard, it appears to me, hence mentioning ffmpeg's cuvid decoding in case some of ffmpeg's code as mentioned in
https://trac.ffmpeg.org/wiki/HWAccelInt ... IDNvDecode
is of interest to you.
Here https://github.com/rdp/ffmpeg-windows-b ... -203004146 mentions
I mention this one in case you were considering a filter to facilitate resizing at various points in scripts and the "libnpp" code may or may not be handy for you.One of the things I'd like to be able to do is hardware-accelerate video resizing during transcoding, which is available using the (nonfree) library libnpp if you have a compatible Nvidia graphics card. I would submit a patch for this, but I could use a little guidance / information on how to do that, because libnpp requires signing up with Nvidia's developer program and then downloading the NVENC SDK (which includes the primitives SDK) in order to get the libraries.
For completeness, also referencing viewtopic.php?f=14&t=559&p=6326#p6326 which links to nvencC code for nvenc encoding.