dav1d 0.9.1 Released With More Optimizations - Particularly For Older CPUs
Dav1d 0.9.1 was released on Sunday as the newest feature update to this leading open-source CPU-based AV1 video decoder.
Dav1d 0.9.1 principally benefits those on older CPUs such as processors supporting SSE2 and SSE4 but not the newer AVX-era Intel/AMD CPUs. There are 10-bit optimizations for SSE4, 10 and 12-bit SSSE3 optimizations, and a variety of other improvements to Dav1d's SSE2/SSE4 Assembly. There is also film grain Arm NEON support and other fixes.
This release is good news for those with pre-AVX Intel/AMD processors for enjoying better AV1 video decode support. Dav1d 0.9.1 can be downloaded from VideoLAN.org.
VideoLAN's Jean-Baptiste Kempf announced that with dav1d 0.9.1, they have "the optimizations are almost finished for all platforms, for all bitdepth (8/10/12) and are now reaching 140,000 lines of code written in assembly...Does this mean that the dav1d project is complete? Well, yes and no. Yes, most of the work for the optimizations are done, and this is good news. However, there is still some major work to be done to improve our threading model; also, there are still some other minor optimizations to be done, and maintenance for new operating systems."
Indeed from my benchmarking so far this weekend the performance on newer Intel/AMD processors is basically a snore... Flat compared to dav1d 0.9, which has already improved a lot in recent time. I don't have any old processors in the racks or otherwise handy that are limited to SSE4, so on short notice I don't have any benchmarks there with this new dav1d release.
See more dav1d benchmarks on OpenBenchmarking.org where you can also run some comparison tests on your own as well.