feat(sinker): Ship 8 — Yuv422p / NV16 RGBA (kernel reuse, no new SIMD)

[uqio]

Tranche 3 of Ship 8 sink-side RGBA. Wiring-only PR — no new SIMD code. The 4:2:2 formats reuse the 4:2:0 kernels from tranches 1 + 2: Yuv422p calls yuv_420_to_rgba_row (already shipped in PR #16), and Nv16 calls nv12_to_rgba_row (already shipped in PR #17). 4:2:2's per-row contract is identical to 4:2:0's — half-width chroma, one pair per Y pair — so the same kernel handles both with no changes.

Scope

Sink-side only, default opaque alpha (0xFF). Per the tracker in docs/color-conversion-functions.md § Ship 8, this is tranche 3 of 7+:

#	Tranche	Formats	Status
1	4:2:0 planar	Yuv420p	✅ shipped (PR #16)
2	4:2:0 semi-planar	Nv12, Nv21	✅ shipped (PR #17)
3	4:2:2 planar + semi-planar	Yuv422p, Nv16	⏳ this PR
4	4:4:4 + 4:4:0	Yuv444p, Nv24, Nv42, Yuv440p	next
5	High-bit-depth 4:2:0	Yuv420p9/10/12/14/16, P010/P012/P016
6	High-bit-depth 4:2:2	Yuv422p9/10/12/14/16, Yuv440p10/12, P210/P212/P216
7	High-bit-depth 4:4:4	Yuv444p9/10/12/14/16, P410/P412/P416

Usage:

use colconv::{
    frame::{Yuv422pFrame, Nv16Frame},
    sinker::MixedSinker,
    yuv::{Yuv422p, Nv16, yuv422p_to, nv16_to},
    ColorMatrix,
};

// Yuv422p (planar): half-width chroma, full-height
let frame = Yuv422pFrame::new(&y_plane, &u_plane, &v_plane, w, h, w, w/2, w/2);
let mut rgba = vec![0u8; (w * h * 4) as usize];
let mut sinker = MixedSinker::<Yuv422p>::new(w as usize, h as usize)
    .with_rgba(&mut rgba)?;
yuv422p_to(&frame, true, ColorMatrix::Bt709, &mut sinker)?;
// rgba[4*i..4*i+3] = R,G,B; rgba[4*i+3] = 0xFF.

MixedSinker::<Nv16> works identically with Nv16Frame + nv16_to (semi-planar interleaved UV).

What's in this PR

Public API

• MixedSinker<Yuv422p>::with_rgba(&mut [u8]) / set_rgba — format-specific impl block, mirrors the Yuv420p pattern.
• MixedSinker<Nv16>::with_rgba(&mut [u8]) / set_rgba — format-specific impl block, mirrors the NV12 pattern.

No new public dispatchers — the existing row::yuv_420_to_rgba_row and row::nv12_to_rgba_row are already on the public surface and handle 4:2:2 byte-identically.

Kernel work

None. Per the existing comments in src/sinker/mixed.rs:

4:2:2 is 4:2:0's vertical-axis twin: same per-row chroma shape (half-width U / V, one pair per Y pair), just one chroma row per Y row instead of one per two. This impl reuses yuv_420_to_rgb_row — no new kernels needed.

That comment is now extended to cover the RGBA path too. Both MixedSinker<Yuv422p>::process and MixedSinker<Nv16>::process invoke the existing _to_rgba_row dispatchers as an independent kernel call (not compose) — same pattern as Yuv420p/NV12 from prior tranches.

Doc updates

• docs/color-conversion-functions.md § Ship 8 tranche tracker — tranche 3 marked ✅ shipped, tranche 4 (Yuv444p / Nv24 / Nv42 / Yuv440p) promoted to "next".
• The compile_fail doctest negative example on MixedSinker::<Yuv420p>::with_rgba moved forward from Nv16 to Yuv444p (the next not-yet-wired format per the tranche tracker).

Tests

+10 lib tests, total 453 (was 443):

Layer	Tests added
Format-level Yuv422p	5: gray-to-gray + opaque alpha, RGB-byte invariant, buffer-too-short, random-YUV SIMD parity, cross-format invariant against Yuv420p RGBA (with solid chroma so 4:2:0 vertical upsample matches 4:2:2's per-row chroma)
Format-level Nv16	5: same shape, with cross-format invariant against NV12 RGBA (duplicated chroma rows so 4:2:0 upsample matches NV16's per-row chroma — mirrors the existing nv16_matches_nv12_mixed_sinker_with_duplicated_chroma RGB-path test)

No per-backend equivalence tests — the underlying kernels (yuv_420_to_rgba_row and nv12_to_rgba_row) are already exercised by the per-backend tests shipped in PR #16 and PR #17. The tranche-3 tests cover the wiring (MixedSinker integration, dispatcher routing); the kernel correctness is established.

Local results (aarch64 macOS): 453 lib tests + 1 doctest pass; wasm32 + x86_64 cross-targets compile clean.

CI matrix coverage is unchanged from PR #17 — the same SSE4.1 / AVX2 / AVX-512 (via SDE) / NEON / wasm32 paths are exercised via the existing per-tier colconv_disable_* rustflags. The new wiring tests run under each tier.

What's deferred

• Tranche 4 — Yuv444p / Nv24 / Nv42 / Yuv440p. The 4:4:4 family is not a kernel reuse — it has its own yuv_444_to_rgb_row family, so this will be a kernel-refactoring tranche similar to PRs feat(sinker): Ship 8 — Yuv420p RGBA output via const-ALPHA template #16 and feat(sinker): Ship 8 — NV12 / NV21 RGBA via const-ALPHA template #17 (const-generic-ALPHA on each backend).
• Tranches 5–7 — high-bit-depth families (Yuv420p9/10/12/14/16, P010/P012/P016, etc.).
• with_rgba_u16 ships in tranches 5–7.
• YUVA source frames (Ship 8b) — independent follow-up.

Test plan

• CI green on test, test-sde-avx512, cross, coverage, clippy, build, miri-* jobs.
• Verify Yuv422p → RGBA + Nv16 → RGBA pipelines end-to-end with real frames.
• cargo doc --lib --no-deps clean (no new doc warnings vs. main).

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds native RGBA output support to additional MixedSinker YUV formats, expanding parity with existing RGB/Luma/HSV paths and validating behavior with new test coverage.

Changes:

• Update MixedSinker<Yuv420p>::with_rgba compile-fail doc example to use a still-not-wired format (Yuv444p).
• Add with_rgba/set_rgba APIs and RGBA row writing to MixedSinker<Yuv422p> and MixedSinker<Nv16>.
• Add comprehensive NV16 and Yuv422p RGBA tests (gray/alpha defaults, RGB-vs-RGBA consistency, buffer-too-short, SIMD-vs-scalar, and cross-format invariants).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

This section header/comment includes an internal milestone label ("Ship 8 PR 3"), which will become stale and isn’t useful context for future readers. Consider renaming to a descriptive, format-focused header (e.g., "Yuv422p RGBA tests").

[Copilot]

The section header/comment mentions an internal milestone ("Ship 8 PR 3"), which is not meaningful long-term and makes the test suite harder to maintain/search. Prefer a timeless, descriptive header (e.g., "NV16 RGBA tests").

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!