Conversation
There was a problem hiding this comment.
Pull request overview
Adds end-to-end support for the P010 (AV_PIX_FMT_P010LE) 10-bit semi-planar YUV 4:2:0 pixel format, including validated frame representation, row-walker API, RGB conversion row primitives, and SIMD backends.
Changes:
- Introduce
P010Frame+ validation (including optional low-bit checking) and a newyuv::P010row-walker (p010_to) withP010Row/P010Sink. - Add P010→RGB row primitives for both
u8RGB and native-depthu16RGB, with SIMD implementations (NEON, SSE4.1, AVX2, AVX-512BW, wasm simd128) plus scalar reference and equivalence tests. - Extend
MixedSinkerto consume P010 rows and optionally populate bothu8RGB andu16RGB outputs; add benchmarks for P010 throughput.
Reviewed changes
Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| src/yuv/p010.rs | New P010 source-format marker, row type, sink trait, and row-walker (p010_to). |
| src/yuv/mod.rs | Register/export the new P010 module and public API. |
| src/frame.rs | Add P010Frame + errors and validation (incl. optional sample low-bit scan). |
| src/sinker/mixed.rs | Add MixedSinker<P010> implementation and tests for P010 behavior. |
| src/row/scalar.rs | Add scalar P010→RGB (u8 and u16) kernels and tests. |
| src/row/mod.rs | Add public P010 row dispatchers with SIMD selection. |
| src/row/arch/neon.rs | NEON P010 SIMD kernels + scalar-equivalence tests. |
| src/row/arch/x86_sse41.rs | SSE4.1 P010 SIMD kernels + scalar-equivalence tests. |
| src/row/arch/x86_avx2.rs | AVX2 P010 SIMD kernels + scalar-equivalence tests. |
| src/row/arch/x86_avx512.rs | AVX-512BW P010 SIMD kernels + scalar-equivalence tests. |
| src/row/arch/wasm_simd128.rs | wasm simd128 P010 SIMD kernels + scalar-equivalence tests. |
| benches/p010_to_rgb.rs | New Criterion benchmarks for P010→RGB (u8 and u16). |
| Cargo.toml | Register the new p010_to_rgb bench target. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Benchmark ResultsBenchmark Results SummaryDate: 2026-04-19 09:29:01 UTC Benchmark Results for macos-aarch64-neonSystem Information
allBenchmark Results for macos-aarch64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-avx2-maxSystem Information
allBenchmark Results for ubuntu-x86_64-defaultSystem Information
allBenchmark Results for ubuntu-x86_64-nativeSystem Information
allBenchmark Results for ubuntu-x86_64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-sse41-maxSystem Information
allBenchmark Results for windows-x86_64-defaultSystem Information
allView detailed resultsDetailed Criterion results have been uploaded as artifacts. Download them from the workflow run to view charts and detailed statistics. |
Benchmark ResultsBenchmark Results SummaryDate: 2026-04-19 09:36:57 UTC Benchmark Results for macos-aarch64-neonSystem Information
allBenchmark Results for macos-aarch64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-avx2-maxSystem Information
allBenchmark Results for ubuntu-x86_64-defaultSystem Information
allBenchmark Results for ubuntu-x86_64-nativeSystem Information
allBenchmark Results for ubuntu-x86_64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-sse41-maxSystem Information
allBenchmark Results for windows-x86_64-defaultSystem Information
allView detailed resultsDetailed Criterion results have been uploaded as artifacts. Download them from the workflow run to view charts and detailed statistics. |
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+1140
to
+1146
| let expected_elements = self.frame_bytes(3)?; | ||
| if buf.len() < expected_elements { | ||
| return Err(MixedSinkerError::RgbU16BufferTooShort { | ||
| expected: expected_elements, | ||
| actual: buf.len(), | ||
| }); | ||
| } |
There was a problem hiding this comment.
MixedSinkerError::RgbU16BufferTooShort’s doc comment says the rgb_u16 buffer is only written by Yuv420p10, but this PR also writes it for P010. Please update that error variant’s documentation to reflect that P010 can populate rgb_u16 as well (and any other formats that do).
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
al8n
changed the title
Benchmark ResultsBenchmark Results SummaryDate: 2026-04-19 09:43:27 UTC Benchmark Results for macos-aarch64-neonSystem Information
allBenchmark Results for macos-aarch64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-avx2-maxSystem Information
allBenchmark Results for ubuntu-x86_64-defaultSystem Information
allBenchmark Results for ubuntu-x86_64-nativeSystem Information
allBenchmark Results for ubuntu-x86_64-scalarSystem Information
allBenchmark Results for ubuntu-x86_64-sse41-maxSystem Information
allBenchmark Results for windows-x86_64-defaultSystem Information
allView detailed resultsDetailed Criterion results have been uploaded as artifacts. Download them from the workflow run to view charts and detailed statistics. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Ships
AV_PIX_FMT_P010LEend-to-end — the HDR hardware-decode keystoneformat emitted by Apple VideoToolbox, VA-API, NVDEC, D3D11VA, and Intel
QSV for 10-bit HEVC / AV1 output. Two output paths: u8 RGB (fast,
downshifts 10→8 in a single Q15 shift) and native-depth u16 RGB
(lossless, low-bit-packed
yuv420p10le-style for downstream HDRconsumers).
Builds on the Ship 2 (yuv420p10) u16 kernel template — same Q15
pipeline, same
range_params_n<10, OUT_BITS>, same chroma-bias math.The new per-backend work is entirely in the u16 semi-planar UV
deinterleave and the
sample >> 6packing extraction (P010's 10active bits live in the high 10 of each u16).
API additions
frame::P010Frame+P010FrameError+ opt-intry_new_checked(low-6-bits-zero sanity check — not a provenance validator; see
docstring).
yuv::{P010, P010Row, P010Sink, p010_to}— marker, row struct,Sink subtrait, row walker.
row::p010_to_rgb_row(u8 out) +row::p010_to_rgb_u16_row(u16native-depth out) dispatchers with SIMD/scalar toggle.
MixedSinker<P010>impl —with_rgb(u8), type-gatedwith_rgb_u16(low-bit-packed),with_luma(>> 8extracts thetop byte of the high-bit-packed sample),
with_hsv(via u8 RGBscratch).
RowSlice::UvHalf10for the semi-planar u16 UV row, newP010FrameError::SampleLowBitsSetwithP010FramePlaneenum.Kernel design
Per-backend u16 UV deinterleave:
vld2q_u16(single instruction — ideal case)._mm_shuffle_epi8per-128-lane split +_mm_unpacklo/hi_epi64combine._mm256_shuffle_epi8+_mm256_permute4x64_epi64+_mm256_permute2x128_si256._mm512_shuffle_epi8+_mm512_permutexvar_epi64+_mm512_permutex2var_epi64(no DQ required).u8x16_swizzle+i8x16_shufflecombine.After deinterleave +
sample >> 6, every backend runs the same Q15pipeline as yuv420p10 — scale, upsample, clamp, narrow, write.
Cross-format equivalence test: same logical 10-bit samples fed
through
yuv420p10(low-packed) andP010(high-packed) producebyte-identical u8 RGB. Verified across every matrix × range.
SIMD backends
All 5 shipped with scalar-equivalence tests across every matrix
(BT.601/709/2020-NCL/SMPTE240M/FCC/YCgCo) × both range modes × tail
widths (18/30/34/1922/1920). NEON has an additional
out-of-range-input adversarial regression (p010-style, random noise,
all-bits-set variants).
CI benchmark (1920px row, scalar → SIMD, ns/iter)
Results from the bench runner across all configured tiers (full
report):
Notes:
colconv_force_scalarrows validate the dispatch gate.4.5–6.2×), reflecting the added u16 UV deinterleave cost that
semi-planar layouts pay over planar — same delta seen in
NV12 vs YUV420p in earlier ships.
write_rgb_u16_8uses 9shuffles + 6 ORs per 8 u16-pixel chunk, a per-pixel cost that
wider tiers absorb better. Room for optimization in a follow-up.
Test plan
cargo test --lib— 160 tests pass (NEON native)cargo build --lib --target x86_64-pc-windows-msvc— cleancargo build --lib --target wasm32-unknown-unknown(simd128) — cleanp010_to_rgb— 2.4–5.7× SIMD speedup across aarch64 + x86_64 + Windows tiers (see table)🤖 Generated with Claude Code