diff --git a/quest/m1/README.md b/quest/m1/README.md index e8bbbfac793..268d2bd2aad 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -59,6 +59,7 @@ the same reason, gates the cut through the - [Vendor and trim the WebRTC stack](/quest/m1/webrtc-vendor.md) - str0m, dimpl, `is`, and sctp-proto live in moq-webrtc with unused code removed, so moq-rtc can take rind - [MoQ on rind](/quest/m1/rind/README.md) - every MoQ crate takes its crypto from rind, a pure-Rust graviola fork, so no build compiles C crypto - [VAAPI encode and decode](/quest/m1/video-vaapi.md) - H.265 encode and decode, pre-generated bindings, and pooled resize surfaces, including the moq-dev/vaapi release that carries them +- [Read a texture back without an encoder](/quest/m1/video-readback.md) - moq-video copies a wgpu texture into packed RGBA with no encoder probe, for the studio's Linux virtual camera ## Related diff --git a/quest/m1/video-readback.md b/quest/m1/video-readback.md new file mode 100644 index 00000000000..af311c27846 --- /dev/null +++ b/quest/m1/video-readback.md @@ -0,0 +1,30 @@ +# [S] Read a texture back without an encoder + +## Goal + +moq-video exposes a public readback that copies a wgpu texture into tightly +packed RGBA rows on the CPU with no encoder behind it, so a consumer that +needs the raw picture reuses the export path's staging buffers, row +unpadding, and mapping lifecycle instead of copying them. A moq-side test +covers it. + +## Plan + +Decided 2026-11-05 (maintainer, on moq-dev/studio's /quest-plan): the +studio's [Linux virtual camera](https://github.com/moq-dev/studio/blob/main/quest/m0/vcam/linux.md) +([moq-dev/studio#9](https://github.com/moq-dev/studio/pull/9)) reads the +program through this readback rather than a copy of it in the studio, and the +work joins m1 because it is additive and that camera has waited on it since +2026-11-02. The studio bumps its pin in +[its own quest](https://github.com/moq-dev/studio/blob/main/quest/m0/pin-readback.md). + +The readback exists: `render/export.rs`'s private `Readback` and its test +`readback_round_trips_and_drops_past_its_depth`. Today it is reachable only +through `Exporter`, whose setup probes for an encoder. Make that type public +and keep one implementation, so `Exporter`'s CPU route still reads through it. + +Recommendation, not settled: mirror `Exporter`'s shape, a non-blocking copy +that drops past its depth and returns a pending handle whose wait yields the +RGBA bytes, named by role under `render`. Refuse a texture whose size or +format does not match, as `Exporter::copy` does. Add a `MIGRATE.md` note +under `## Unreleased`. diff --git a/quest/m3/README.md b/quest/m3/README.md index 14548b8c020..7e4a066f378 100644 --- a/quest/m3/README.md +++ b/quest/m3/README.md @@ -133,6 +133,7 @@ federation) land. - [Compile time audit](/quest/m3/audit-compile-time.md) - build-time hotspots are measured against the m1 baselines, starting with the transport share now that moq-net compiles once, and the worst are removed, made cheaper, or planned - [Lint override and deprecation audit](/quest/m3/audit-lints.md) - repeat the suppression and deprecation audit after the release ships, starting with clippy on the macOS and Windows targets - [Clean ALSA playback onset](/quest/m3/audio-alsa-onset.md) - no onset restarts or clicks over PulseAudio +- [Read a bus source's own samples](/quest/m3/audio-strip-tap.md) - a tap reads one source's samples before its delay with their bus-clock positions, for the studio's clap sync - [Interop audit](/quest/m3/audit-interop.md) - our interop runner entries match what we ship, and every red interop cell with a moq.dev side is fixed, reported, or noted as down or unsupported - [moq-hls exposes each rendition's bound source](/quest/m3/hls-rendition-source.md) - playlists and segments report the source path and instance they read, so a cache can key on the source - [The memory bench finishes at 1 MiB frames](/quest/m3/memory-bench-hang.md) - the allocation mode's park at `size=1048576` is root-caused and fixed diff --git a/quest/m3/audio-strip-tap.md b/quest/m3/audio-strip-tap.md new file mode 100644 index 00000000000..31108b03f1f --- /dev/null +++ b/quest/m3/audio-strip-tap.md @@ -0,0 +1,34 @@ +# [S] Read a bus source's own samples + +## Goal + +A `moq_audio::bus` caller reads one source's samples as they enter the bus, +before its strip's delay, as `Block`s stamped with their position on the bus +clock, so it can time an event in that source against the clock. A tap that +falls behind skips, as a mix's `Reader` does, and never stalls the bus or +another reader. + +## Plan + +Decided 2026-11-05 (maintainer, on moq-dev/studio's /quest-plan): + +- ✅ **A sample tap.** moq-audio exposes the samples and their positions; + the consumer detects whatever event it is timing. + Rejected: an onset detector in moq-audio, which would put clap-specific + DSP and tuning here. +- ✅ **m3**: additive, with a named consumer that is not urgent. + +The consumer is the studio's +[clap-to-measure sync delay](https://github.com/moq-dev/studio/blob/main/quest/m0/sync-clap.md): +a clap seen by a camera and heard by a microphone measures how far the +microphone leads, and sets its `Strip::set_delay`. The strip's level meter, +read once a video frame, is too coarse to time that. The tap sits before the +delay because the delay is what is being measured; measuring after it would +include the current setting. The studio bumps its pin in +[its own quest](https://github.com/moq-dev/studio/blob/main/quest/m0/pin-audio-tap.md). + +Recommendation, not settled: reuse `Reader` and `Block`, whose `position` is +already the bus clock, rather than a new block type. Taking the samples +before gain and mute too lets a muted microphone be measured. A source with +no tap costs nothing extra. Test with an impulse written at a known point: the +tap reports it at the expected bus position whatever the strip's delay.