On 9/8/2018 5:38 PM, Paul B Mahol wrote:
I'll probably remove the interp_algo from this before committing, hard
coding it to nearest neighbor. I'm unable to see any difference between
them for perfect 90DEG angles except that NN is easily 10 times faster
than the current default Cubic.

Perhaps interpolation is useful for other pixel format where vertical
and horizontal subsampling are not same.

This only supports yuv420p and yuv444p so I don't think it's an issue. It's impossible for such formats to end up in a CUDA frame to begin with, since nothing supports putting them in there.

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to