FFmpeg
FFmpeg copied to clipboard
ffvvc
AVX2 ALF filter
| clip | before(fps) | after(fps) | delta |
|---|---|---|---|
| Tango2_3840x2160_60_10_420_27_LD.266 | 56 | 115 | 105.36% |
| RitualDance_1920x1080_60_10_420_32_LD.266 | 272 | 481 | 76.83% |
| RitualDance_1920x1080_60_10_420_37_RA.266 | 303 | 426 | 40.59% |
@nuomi2021 Looks like there is a very huge improvement on the i9 14900k platform.
i9 14900k platform.
💕
Yes, ALF is the most time-consuming part of the vvc decoder.
@nuomi2021 Should we need to pass the new fuzz test?
@QSXW Try with https://github.com/ffvvc/tests/pull/34. Sorry, did not update the failing bitstreams after 3d7372594409341687e35824883a18049edf996b and it seems there are some bitstreams which fail on the GitHub actions runners despite passing on higher-powered machines. These same bitstreams also failed here so they do not seem related to your changes.
