On 19/09/18 02:33, myp...@gmail.com wrote:> > VA-API encoder can support multiple-frame reference too.
That's loosely the intent, though it probably needs more information about similarities between frames for the feature to be of non-placebo use. In terms of the common implementation it should be possible to make an alternative pick_next() function which creates the set of references in a different way and have it work for all codecs (or at least H.264/5, VP9 and in future AV1; MPEG-2 and VP8 have constraints making them less general), though the codec-specific parts do still contain some assumptions about structure which would need to be ironed out for that to actually work. I think the information to make it useful could either come from somewhere external (e.g. a filter doing some sort of frame-similarity test and passing that information on), or it could be internal by running the encoder itself multiple times in different configurations and looking at the resulting output. Any thoughts you have on how this should work are welcome. On 19/09/18 04:20, myp...@gmail.com wrote: > On Wed, Sep 19, 2018 at 6:35 AM Mark Thompson <s...@jkqxz.net> wrote: >> >> --- >> libavcodec/vaapi_encode_h264.c | 13 +++++++------ >> 1 file changed, 7 insertions(+), 6 deletions(-) >> >> ... > Missing “slices” option in doc/encoder VA-API part, I think. other part LGTM Yep, added in new version. Thanks, - Mark _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
