aiongg
0a881266c8
* Tests: one integration-test binary per crate Each crate with several tests/*.rs files keeps them as modules of tests/it/, so cargo links one test executable per crate instead of one per file. Files that change process-wide state or time a workload stay binaries of their own. * Profiles: line tables for workspace crates, no debug info for dependencies * Windows ARM64: nightly and on release pushes instead of every push to main * xtask ci: clippy and tests in the dev profile The release profile's thin LTO and single codegen unit made linking the test binaries take most of the gate. The codec and audio crates stay at opt-level 3 in the dev profile, so their tests run at release speed. * xtask ci: nextest, doctests, no examples, one wasm check Tests run with cargo nextest when it is installed (cargo test otherwise) and build with --lib --bins --tests, so examples aren't linked; doctests run separately. The wasm check keeps going past the first failing crate. VP8 fixture generation locks across processes, since nextest runs each test in its own. * VP9 synthetic fixtures: write aside and rename A test in another process could read a fixture ffmpeg was still writing. * Fuzz tests: a third of each sweep by default, all of it with FILMCRAFT_FUZZ_ITERS Every seed, mutation kind and thread count still runs. Header bytes each get one overwrite value and two bit flips in rotation, so every value and bit index still occurs across the region. FILMCRAFT_FUZZ_ITERS=N runs the whole sweep plus N random rounds per seed, as before. * Platform fallback tests: one failure point per hybrid path The failure-anywhere test injected a failure at four points around every restart point; the hybrid's paths differ only in the first GOP, around the second and last restart points, at the end and never. The replay test needs the first two GOPs. * Export tests: shorter ranges with the same checks The bitrate test measures 0.5–1.5 s, where each preset reaches its own target (on 0–3 s the simple demo left two presets at the same rate). The every-preset test exports three frames per preset. * Audio tests: effects in parallel, a table-driven FFT for the Opus quality metric The extreme-settings test keeps both settings and the 40 s tail for every effect, one thread each. The Opus quality metric's FFT builds its twiddles once and allocates nothing per transform (same algorithm and summation order). * CI: dev-profile gate with nextest xtask ci no longer builds with --release, so the release-profile overrides go; nextest is installed for the test step. * CI: apt packages through the cached install-apt-packages action * Review fixes: fuzz tests with a panic hook in their own binaries, loud fixture lock The fuzz tests of av1, h264, hevc, mpeg2v, vp8 and vp9 count panics through a process-wide hook, so a panic in another test of the same process would fail them; they are binaries of their own again. The VP8 fixture lock fails the test instead of skipping it silently, and the full-sweep doc comments say how many truncation points the full run samples. * Fuzz tests: rotate the default subset per run (FILMCRAFT_FUZZ_SEED replays one) * platform tests: decode-switch tests in their own binaries, one NVENC lock
4.2 KiB
filmcraft-vp8
Clean-room, pure-Rust (no unsafe) VP8 decoder, implemented from RFC 6386, VP8 Data Format and
Decoding Guide, sections 1-19. It is an L0 crate: it depends only on thiserror, and builds for
wasm32-unknown-unknown.
The constant tables (coefficient and mode probabilities and their update probabilities, the
subblock mode contexts, quantizer lookups, MV probabilities, interpolation filters...) were
extracted from the text of the RFC by tools/extract_tables.py into
src/spec_tables.rs; the script cuts the text at section 20, so nothing is read from the
reference decoder attached to the RFC. No code was taken or ported from libvpx, FFmpeg or any
other decoder; ffmpeg (with libvpx) and vpxenc are used only as external fixture generators and
test oracles.
Status
| Area | Status |
|---|---|
| Frame tag, key frame header, frame header (segmentation, loop filter type / level / sharpness, deltas, partitions, quantizer indices, reference updates and sign bias, probability updates) | done |
Boolean decoder, tree decoding, probability persistence across frames and refresh_entropy_probs |
done |
| Intra prediction: 16x16 and chroma modes, the ten 4x4 subblock modes, key frame subblock mode contexts, edge rules | done |
| Inter prediction: last / golden / altref with sign bias, NEAREST / NEAR / ZERO / NEW / SPLIT (all four partitionings), MV decoding and clamping, six-tap and bilinear filters, full-pixel chroma (bitstream versions 0-3) | done |
| Coefficient tokens (1-8 partitions), dequantization (frame and segment quantizers), inverse WHT and DCT | done |
| Normal and simple loop filters, per-macroblock levels from segments and reference / mode deltas | done |
| Golden / altref copies, hidden (altref) frames, frame size changes at key frames | done |
| Damaged input: errors (never panics), decoding resumes at the next key frame | done |
| Upscaling hints of the key frame header (9.1) | not applied (reported in KeyframeInfo) |
Bit-exact with libvpx on the 62 libvpx VP8 test vectors checked (vp80-00-comprehensive-001 to
018 and the vp80-01 to vp80-06 intra, inter, segmentation, partition, sharpness and small
size vectors; tests/it/vectors.rs checks them when they are present) and on the generated fixtures
of tests/it/conformance.rs.
API
use filmcraft_vp8::Decoder;
let mut dec = Decoder::new();
for (pts, frame) in frames { // one IVF frame / WebM block / MP4 sample each
let skippable = dec.is_disposable(frame); // could this frame be left out (catching up)?
if let Some(pic) = dec.decode(frame, pts)? {
// pic.width / pic.height, pic.chroma_width / chroma_height,
// pic.y / pic.u / pic.v: tight 8-bit 4:2:0 planes, pic.pts, pic.key,
// pic.color_space (0: BT.601 YUV)
}
// None: a hidden (altref) frame
}
dec.reset(); // forget the reference frames (seek): decoding resumes at the next key frame
let stats = dec.stats(); // frames, key frames, hidden frames
// Random access points
let key: bool = filmcraft_vp8::is_keyframe(frame);
let info: Option<filmcraft_vp8::KeyframeInfo> = filmcraft_vp8::keyframe_info(frame);
filmcraft-codecs wires the decoder into Matroska / WebM (V_VP8, including the alpha layer of
transparent WebM in BlockAdditional) and MP4 (vp08). VP8 has no colour signalling: pictures are
BT.601, studio swing (RFC 6386 9.2), unless the container's Colour says otherwise. VP8 has no
superframes either: a hidden altref frame is a block of its own, and the WebM reader takes a
request for its time as a request for the shown frame that follows it.
Tests and tools
cargo test -p filmcraft-vp8: unit tests, bit-exactness against libvpx on generated fixtures (target/fixtures/vp8-tests, skipped without ffmpeg / libvpx), robustness (corrupted, truncated, reordered and random input), and the libvpx test vectors when present intarget/fixtures/vp8/conformance(or$VP8_TEST_VECTORS).cargo run --release -p filmcraft-vp8 --example vp8dec in.ivf out.yuv: decode to raw I420.cargo run -p filmcraft-vp8 --example fuzz -- <iterations> <seed> files.ivf...: decode randomly damaged variants of IVF files (in a debug build, so overflow panics too) and report any panic.