On 14/11/2020 17:32, Timo Rothenpieler wrote:
On 14.11.2020 16:08, James Almer wrote:
On 11/14/2020 12:01 PM, Timo Rothenpieler wrote:
I still strongly dislike this whole thing.
The same logic is ran essentially twice.
While it's applying the same method it is not doing the same thing, because
different inputs are being used.
An ideal solution would would only run this logic once, in the AV1
decoder/parser.
No it wouldn't.
You can't do it in the parser, because the parser doesn't know which frames
have been successfully decoded. The parser still has to run the
standard-dictated logic, though, because it needs to know whether the
skip_mode_present syntax element is present.
The other hwaccel implementations have made the decision to push it into the
driver, which definitely knows which frames have decoded successfully - I'm
pretty sure that's the right answer for a correct result.
Nvdec forcing us to guess in the wrapper layer which doesn't have complete
information is a problem, but we can at least make a better guess than the
parser did.
- Mark
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
