2018-06-17 22:44 GMT+02:00, Mark Thompson <[email protected]>: > On 17/06/18 21:30, Carl Eugen Hoyos wrote: >> 2018-06-17 22:00 GMT+02:00, Mark Thompson <[email protected]>: >> >>> Intel devices do not support MPEG-4 part 2 at all. If you are using >>> the Mesa driver on an AMD device, some MPEG-4 part 2 streams >>> may be supported if you set the environment variable >>> VAAPI_MPEG4_ENABLED to 1, but it's not enabled by default >>> because the implementation is incomplete due to API constraints. >> >> I know this mail isn't helpful but reading above and remembering how >> nice vdpau worked (already) many years ago and remembering how >> people said at the time that vaapi is much superior over vdpau I still >> wonder if we (FFmpeg) shouldn't have fought vaapi much stronger... > > I doubt it would be particularly hard to make it work. Since Intel didn't > implement MPEG-4 part 2 on their supported devices they just cobbled > together some set of stream/frame properties which looked sufficient for > decoding and made that the API. When AMD / Mesa tried to implement > it they made some of it work but found that the API was incomplete, and > then they gave up. If you (or anyone else, including AMD) wants it to > work then it would be straightforward (if somewhat tedious) to add the > necessary properties to VAAPI / libva to get to the point that it matches > the working VDPAU behaviour.
Sadly that would be far above my capabilities. Again, sorry for the rant, I am happy for everybody who can use vaapi. Carl Eugen _______________________________________________ ffmpeg-user mailing list [email protected] http://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email [email protected] with subject "unsubscribe".
