Files

T

marfrit 356e446a49 Cycle 3 (MC interpolation) closure: M1'''=100%, R'''=0.067 RED, M4=-19.5%

Third daedalus-fourier kernel — VP9 8-tap regular subpel filter,
horizontal direction, 8-wide output. Multiply-heavy by design to
stress V3D's no-DP4A deficit. Full cycle Phase 1-7 + M4'''.

Phase 5''' second-model review delivered cleanly — caught 1 RED
bug pre-implementation (src_off off-by-3 indexing convention) and
2 YELLOW gaps (assert MUST language, shaderdb filter-LUT gate).
Without the review, M1''' would have failed silently on first run
with cryptic "high-index source pixels wrong" symptoms.

Phase 6 v1 first-light: M1''' 100.0000% bit-exact (65536/65536
blocks across all 16 mx phases). Phase 5''' filter-LUT prediction
materialised exactly: 197 uniforms (gate was 144), 2 threads (down
from cycle-2's 4 due to register pressure).

Performance:

  M2''' = 1.413 Mblock/s     (707.9 ns/block)
  M3''' = 20.997 Mblock/s    (NEON baseline phase3)
  R'''  = 0.067              (RED band — structural mismatch)
  shaderdb: 488 inst, 2 threads, 197 uniforms, 25 max-temps, 0 spills

M4''' concurrent matrix (8s windows):

  NEON 1-core           14.479 Mblock/s
  NEON 4-core           15.248 Mblock/s   <- baseline (compute-bound,
                                              not bandwidth-saturated
                                              like cycles 1+2!)
  QPU only               1.380 Mblock/s
  MIXED NEON-3 + QPU    12.277 Mblock/s   <- -19.5% (FAIL gate)
  MIXED NEON-4 + QPU    12.158 Mblock/s   <- -20.3%

NEW cross-cycle finding (Phase 9 lesson 2): compute-bound CPU
workloads make the QPU-offload story collapse. Cycles 1+2 were
bandwidth-saturated (4-core scaling 0.56-0.82x of 1-core), so
freeing a CPU core via QPU offload added throughput. Cycle 3 MC
is compute-bound (4-core scaling 1.05x of 1-core — near-linear),
no free cycles to free. QPU contribution (0.45 Mblock/s in
contention) doesn't compensate for losing 1 NEON core delivering
~3.8 Mblock/s.

But 30fps@1080p floor: PASS in every config (1.4x to 15.7x
isolation margin). Per project_30fps_floor_is_fine.md, user-facing
test never fails — daily YouTube playback works fine on any CPU/QPU
split.

DEPLOYMENT RECIPE for higgs (cycle 3 confirmed split):

  IDCT (k1)  -> QPU   (R=0.92, +7% mixed, frees CPU core)
  LPF  (k2)  -> QPU   (R=0.41, +7% mixed, frees CPU core)
  MC   (k3)  -> CPU   (R=0.067, -19.5% mixed — stays on CPU)
  Entropy    -> CPU   (structurally serial)

Mixed-substrate deployment, not "QPU does everything". Realistic for
higgs: entropy + MC on 2-3 ARM cores; IDCT + LPF dispatched to QPU
concurrently; 1-2 ARM cores left for vscode etc.

New artifacts:
- src/v3d_mc_8h.comp               — GLSL kernel
- tests/vp9_mc_ref.c               — standalone C ref (REGULAR filter
                                     embedded; clean transcription)
- tests/bench_neon_mc.c            — M1'''_c + M3''' bench
- tests/bench_v3d_mc.c             — M1''' + M2''' bench with contract
                                     asserts + 30fps margin display
- tests/bench_concurrent_mc.c      — M4''' pthread bench
- external/ffmpeg-snapshot/libavcodec/aarch64/vp9mc_neon.S    (vendored)
- external/ffmpeg-snapshot/libavcodec/vp9_subpel_filters_table.c
                                     (hand-extracted; provides
                                      ff_vp9_subpel_filters symbol
                                      without dragging in full vp9dsp.c)
- docs/k3_mc_phase{1,2,3,4,5,7}.md — full cycle documentation

Memory updates: project_30fps_floor_is_fine.md (user's 30fps target
recalibration), MEMORY.md index updated.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-18 12:51:43 +00:00

4.2 KiB

Raw Blame History

FFmpeg source snapshot

Verbatim subset of FFmpeg source pinned for use as reference implementations of the VP9 8×8 inverse DCT (Phase 1 target of daedalus-fourier). See ../../docs/phase2.md §2 and §5 for the rationale.

Upstream pin

Repository: https://github.com/FFmpeg/FFmpeg
Tag: n7.1.3 (matches libavcodec61 8:7.1.3-0+deb13u1+rpt1 shipping in Debian Trixie on the dev host hertz)
Annotated tag object: 0a9a757e96fdf053697084bbd1f620edeac9d084
Commit object (tag target): f46e514491172d15bd74b4abb1814cd2f05a763e
Snapshot fetched: 2026-05-18 (UTC), via https://raw.githubusercontent.com/FFmpeg/FFmpeg/n7.1.3/<path>

Files in this snapshot

All files are byte-for-byte copies of the upstream source at the tagged commit, no modifications.

Path	Lines	Bytes	SHA-256
`libavcodec/vp9dsp_template.c`	2578	89045	`41b21f667a6c497b620aa1637d8269badc45d1ac7e621d694441c5bf39356e4f`
`libavcodec/aarch64/vp9itxfm_neon.S`	1580	63534	`82ee3ceed4735c63576bafdcee28e2215652743ade55a9eab46a16d9530369f6`
`libavcodec/aarch64/vp9lpf_neon.S`	1334	—	`384e49e7a6e838d9e38aedc00838ed4aebfa6c5bdb343ecaf23ef639bc10fbb7`
`libavcodec/aarch64/vp9mc_neon.S`	665	—	`6b1d50f9821742584fdd47758057f810644aff3a008faaa774ff5b9cac4d1fef`
`libavcodec/vp9_subpel_filters_table.c`	—	—	hand-extracted from `libavcodec/vp9dsp.c` at same n7.1.3 pin — provides `ff_vp9_subpel_filters` for `vp9mc_neon.S` to link against without dragging in vp9dsp.c's full init machinery
`libavcodec/aarch64/neon.S`	173	7496	`72d36ce6c3fcc5e53de869cfe10fda16225ebe580c32891bccc240a30a85a538`
`libavutil/aarch64/asm.S`	260	8069	`c0d03143b1bc5a9e358222d08d2d449d595271844fe7a3dc23bffb91abe8b0e3`
`COPYING.LGPLv2.1`	502	—	`b634ab5640e258563c536e658cad87080553df6f34f62269a21d554844e58bfe`

Verify with:

( cd external/ffmpeg-snapshot && sha256sum -c <<'EOF'
41b21f667a6c497b620aa1637d8269badc45d1ac7e621d694441c5bf39356e4f  libavcodec/vp9dsp_template.c
82ee3ceed4735c63576bafdcee28e2215652743ade55a9eab46a16d9530369f6  libavcodec/aarch64/vp9itxfm_neon.S
72d36ce6c3fcc5e53de869cfe10fda16225ebe580c32891bccc240a30a85a538  libavcodec/aarch64/neon.S
c0d03143b1bc5a9e358222d08d2d449d595271844fe7a3dc23bffb91abe8b0e3  libavutil/aarch64/asm.S
b634ab5640e258563c536e658cad87080553df6f34f62269a21d554844e58bfe  COPYING.LGPLv2.1
EOF
)

License

LGPL-2.1-or-later. See COPYING.LGPLv2.1. Original copyright holders include the FFmpeg authors and Google Inc. (2016) for the aarch64 NEON paths. The snapshot inherits FFmpeg's license in full.

Why each file is in this snapshot

libavcodec/vp9dsp_template.c — contains idct_idct_8x8_add_c, the bit-exact C reference for the Phase 1 kernel under test (M1).
libavcodec/aarch64/vp9itxfm_neon.S — contains ff_vp9_idct_idct_8x8_add_neon, the NEON throughput baseline (M3). Also defines idct8, dmbutterfly0, dmbutterfly, dmbutterfly_l, butterfly_8h, and the idct_coeffs constant table.
libavcodec/aarch64/neon.S — defines transpose_8x8H used by vp9itxfm_neon.S.
libavutil/aarch64/asm.S — defines function, endfunc, movrel, const, endconst, and other assembly preamble macros required to assemble the above NEON files.

Re-vendoring procedure

If the upstream pin needs to change (e.g., hertz updates to a newer libavcodec):

TAG=nX.Y.Z
BASE=https://raw.githubusercontent.com/FFmpeg/FFmpeg/$TAG
cd external/ffmpeg-snapshot
for f in libavcodec/vp9dsp_template.c \
         libavcodec/aarch64/vp9itxfm_neon.S \
         libavcodec/aarch64/neon.S \
         libavutil/aarch64/asm.S \
         COPYING.LGPLv2.1; do
  curl -sSf -o "$f" "$BASE/$f"
done
sha256sum libavcodec/vp9dsp_template.c \
          libavcodec/aarch64/vp9itxfm_neon.S \
          libavcodec/aarch64/neon.S \
          libavutil/aarch64/asm.S \
          COPYING.LGPLv2.1
# update this PROVENANCE.md with the new tag, commit hash, and hashes

After re-vendoring, re-run the bit-exact gate (M1) and throughput baseline (M3) — both can shift across FFmpeg versions even when the VP9 spec doesn't change (e.g., NEON micro-optimizations).

4.2 KiB Raw Blame History Unescape Escape