Commit Graph
4 Commits
Author SHA1 Message Date
Brandon ThomasandClaude Opus 5.5 842801de41 M4.10: VP9 frame threading (band-published references, per-frame post stage) and draft mode
Frames are published one superblock row at a time once loop filtered; a post stage per frame
(own thread) assembles bands from the tile strips, filters them in raster order and builds the
output picture, while the next frame's tile columns decode and wait per band for the reference
rows they read. Bit-exact with 1, 3 and all threads. Draft mode skips the loop filter of frames
that refresh no reference slot. av1dec / hevcdec bench knobs and VP9 / AV1 4K playback scenarios
in the bench.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-03 05:34:39 -04:00
Brandon ThomasandClaude Opus 5.5 0d1f21d393 M9.8: VP9 decode speedup — lane-parallel loop filter (i16/i32 lanes), banded parallel loop filter, vectorizable MC filters, per-size transform buffers, buffer pools, parallel strip assembly and picture output
1080p 8-bit bench: 55 -> 98 fps single-thread CPU time; 45 -> 115 fps wall with 14 threads
(best of 4 on a heavily loaded machine). Still bit-exact against the ffmpeg oracle on the
whole fixture matrix (1 thread and thread pool). Fixture generation uses per-call unique
temporary names (concurrent tests generating the same fixture raced on rename).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-01 00:21:03 -04:00
Brandon ThomasandClaude Opus 5.5 5e9d7d904e M9.8: VP9 loop filter speedup (gathered samples, sliding wide filter), perf test, vp9dec example, README
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 22:46:13 -04:00
Brandon ThomasandClaude Opus 5.5 804eb14fa9 M9.8: VP9 fixture matrix with backward adaptation / alt-ref / tile rows, synthetic scaled-reference, intra-only and show_existing streams, robustness tests
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 22:37:46 -04:00