Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
39 changes: 8 additions & 31 deletions Cargo.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

6 changes: 3 additions & 3 deletions Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -153,8 +153,8 @@ moq-tokio = { version = "0.19.23", path = "rs/moq-tokio", default-features = fal
# Default features off on moq-transcode and moq-video so each workspace consumer
# chooses native codecs and rendering explicitly. OpenH264 is opt-in everywhere:
# each consumer forwards an `openh264` feature that release builds turn on.
# VAAPI is opt-in everywhere; its decoder is hardware-validated, while its
# encoder is not yet.
# NVIDIA and VAAPI dlopen their drivers and need nothing on the build host, so
# the consumers that build video turn both on.
moq-transcode = { version = "0.1.10", path = "rs/moq-transcode", default-features = false }
# default-features off (the `quic` backend) so the consumer opts into the QUIC
# stack the io_uring path compiles; cargo features are additive, so a default-on
Expand All @@ -166,7 +166,7 @@ moq-uring = { version = "0.0.12", path = "rs/moq-uring", default-features = fals
moq-v4l = { version = "0.1.0", path = "rs/moq-v4l" }
# Standalone crate (moq-dev/vaapi); vendored from cros-libva + cros-codecs.
# dlopen's libva at runtime (no libva-dev at build, no NEEDED libva in the binary).
moq-vaapi = "0.1.1"
moq-vaapi = "0.2.0"
# `default-features = false` is here for `capture` (V4L2, libclang), which the
# cross-compiled Python, Swift, Kotlin, and Go builds have no use for: adding
# `features = ["capture"]` to a consumer that ships in those bindings pulls the
Expand Down
8 changes: 4 additions & 4 deletions doc/lib/rs/moq-video.md
Original file line number Diff line number Diff line change
Expand Up @@ -40,10 +40,10 @@ change bitrate live follow the connection's send estimate while someone
watches, and return to the configured bitrate while nobody does. Where the platform
allows it, frames stay on the GPU between capture, decode, encode, and render.

Opt-in features: `capture`, `vaapi` (Intel and AMD on Linux; the build needs
libclang), `v4l2` (ARM SoC codecs such as a Raspberry Pi's), `pipewire` (Linux
portal capture and PipeWire cameras; links libpipewire), and `vpx` (links
libvpx).
Opt-in features: `capture`, `v4l2` (ARM SoC codecs such as a Raspberry Pi's),
`pipewire` (Linux portal capture and PipeWire cameras; links libpipewire), and
`vpx` (links libvpx). `vaapi` (Intel and AMD on Linux) is on by default beside
`nvidia`, since neither needs anything on the build host.

On Wayland, the `pipewire` feature lets Rust callers request a screen or
window through the system picker with `capture::Source::Portal`. Keep the
Expand Down
5 changes: 0 additions & 5 deletions nix/overlay.nix
Original file line number Diff line number Diff line change
Expand Up @@ -142,11 +142,6 @@ let
cargoExtraArgs = "-p moq-c --features openh264";
nativeBuildInputs = with final; [
pkg-config
# moq-c is the only nix-built package that pulls moq-video, and its `vaapi`
# feature brings moq-vaapi, whose build.rs runs bindgen over its vendored
# libva headers. Sets LIBCLANG_PATH + BINDGEN_EXTRA_CLANG_ARGS so it finds
# libclang and the libc headers, same as the devShell in flake.nix.
rustPlatform.bindgenHook
];

# moq-c is a staticlib; crane's default install phase only handles
Expand Down
3 changes: 1 addition & 2 deletions quest/m1/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ Placements decided in the re-cut: read-backfill (a JS fix that matches Rust),
buffered-ascending (lite-07 already carries the order field), webrtc-vendor
(internal, though it requires the m2 rename), and the crates.io trusted
publisher and OIDC quests (setup after the cut's first publishes) are
additive. [video-vaapi](/quest/m1/video-vaapi.md) moved up from later work
additive. video-vaapi moved up from later work
because a downstream iGPU transcode trial needs it now.

Held here by open claims: a quest's branch is its path, so a claimed quest
Expand Down Expand Up @@ -49,7 +49,6 @@ moves what remains of them once their claims merge.
- [Session policy lives in moq-net](/quest/m1/session-policy/README.md) - accept and connect deadlines, admission, window widening and reconnect live once in moq-net; moq-tokio and moq-relay's io_uring path call them
- [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-webrtc 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
- [VA-API encode is correct on radeonsi](/quest/m1/vaapi-radeonsi.md) - AMD's VA-API H.264 encode decodes, or the node is refused for OpenH264
- [Tests run on a mocked clock](/quest/m1/mock-clock-tests/README.md) - about 300 tests stop proving nothing happened by waiting, sleeping to settle, or asserting on elapsed time
- [moq-mux reads H.265 SPS without scuffle](/quest/m1/h265-sps-reader.md) - moq-mux's own SPS reader replaces scuffle-h265 and scuffle-expgolomb, and the fuzz regressions still pass
Expand Down
65 changes: 0 additions & 65 deletions quest/m1/video-vaapi.md

This file was deleted.

1 change: 0 additions & 1 deletion quest/m3/obs-moq-video/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -35,4 +35,3 @@ Decided in the 2026-09-30 audit: the Windows and Linux GPU decode paths and the
- [AudioToolbox decode](/quest/m3/audio-decode-audiotoolbox.md) - HE-AAC and multichannel AAC reach the OBS source on macOS through moq-ffi
- [AudioToolbox encode](/quest/m5/audio-encode-audiotoolbox.md) - native AAC encode reaches the OBS encoder adapter on macOS through moq-ffi
- [Linux GPU input](/quest/m5/obs-linux-gpu.md) - allocation-export feasibility and its dependent implementation are deferred
- [VAAPI encode and decode](/quest/m1/video-vaapi.md) - owns Linux backend decode/import capabilities; reconcile its older dependency assumptions against current code
3 changes: 1 addition & 2 deletions quest/m3/vaapi-vulkan-import.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ the same type reaches CUDA, Vulkan Video, and VA-API.
Decided 2026-10-05 in a `/quest-plan` interview (paper trail in the PR that
added this quest): Intel uses VA-API, since ANV on Arrow Lake has no Vulkan
encode on Mesa 26.0.8 (it is in 26.2 behind `ANV_DEBUG`). H.264 is enough for
the proof; H.265 on VA-API belongs to [VAAPI encode and decode](/quest/m1/video-vaapi.md).
the proof; H.265 on VA-API is already on main.

Today the VA-API encoder takes a `Surface::DmaBuf` and waits for the producer
through an implicit-fence poll with a 500 ms timeout (`DMA_BUF_FENCE_TIMEOUT`
Expand Down Expand Up @@ -70,7 +70,6 @@ Public API: none; the surface landed in #4975. Wire: none.
## Related

- [Vulkan Video encode on AMD](/quest/m3/vulkan-encode.md) - the AMD half of the same proof
- [VAAPI encode and decode](/quest/m1/video-vaapi.md) - H.265 encode on the same backend
- [GPU capacity and health](/quest/m3/gpu-health.md) - reports per device keyed by the same render node `dev_t`
- [Linux OBS GPU input](/quest/m5/obs-linux-gpu.md) - a DMA-BUF producer that needs the same explicit sync
- [Video hardware validation](/quest/m5/video-hardware.md) - VA-API so far ran only on Meteor Lake
2 changes: 2 additions & 0 deletions quest/m4/vaapi-render-tests.md
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,8 @@ return early without a device today (`vaapi_h264_round_trip` and
`native_frames_still_answer_into_i420` in
`rs/moq-video/src/decode/backend/vaapi.rs`) and moq-transcode's
`vaapi_decoder_available` tests, which hosted CI reports as passed.
The ignored `vaapi_1080p_ladder_stays_on_the_gpu` encode test in
`rs/moq-video/src/encode/backend/vaapi.rs` runs there too.

## Plan

Expand Down
4 changes: 0 additions & 4 deletions quest/m5/obs-decode-linux.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,3 @@ At least one supported Linux hardware decoder delivers frames to OBS without CPU
## Required

- [Video source replacement](/quest/m3/obs-moq-video/source.md) - native frame contract and fallback lifecycle

## Related

- [VAAPI encode and decode](/quest/m1/video-vaapi.md) - owns remaining VAAPI decode and native surface gaps; avoid a duplicate backend implementation
1 change: 0 additions & 1 deletion quest/m5/obs-linux-gpu.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,4 +24,3 @@ the bottleneck.

- [Video hardware validation](/quest/m5/video-hardware.md) - native input and encoder acceptance need hardware evidence
- [VA-API encodes an external Vulkan image](/quest/m3/vaapi-vulkan-import.md) - explicit sync for an external DMA-BUF, and the public `DmaBuf` constructor from #4975
- [VAAPI encode and decode](/quest/m1/video-vaapi.md) - H.265 and checked-in bindings. The DMA-BUF encoder import this quest needs is already on main (moq-vaapi 0.1.1).
5 changes: 2 additions & 3 deletions quest/m5/video-codec-coverage.md
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,8 @@ and a portable software AV1 fallback such as
OS, extension, hardware, format, and fixture requirements; a browser playing
a codec is not proof that our native backend can open it.

NVIDIA AV1 encode/10-bit and VAAPI expansion already have quests. Keep those
owners. Software VP8/VP9 decode is libvpx behind moq-video's `vpx` feature;
NVIDIA AV1 encode/10-bit already has quests; keep those owners. VA-API gaps,
such as HEVC Main 10 decode, have no owner and are in scope here. Software VP8/VP9 decode is libvpx behind moq-video's `vpx` feature;
hardware VP8/VP9 decode is in scope here. Windows NVIDIA support needs a demonstrated advantage over
the existing native path. If revisiting software AV1 encoding, measure the
target real-time workload and build cost rather than assuming all presets or
Expand All @@ -32,5 +32,4 @@ Decided 2026-10-08: moved to m3. No consumer has asked for a missing codec.
## Related

- [NVIDIA formats](/quest/m5/2147-moq-video-10-bit-hevc-and-av1-support-in-the-nvidia-codec.md) - 10-bit HEVC scope; AV1 encode moved to [NVENC AV1](/quest/m5/nvenc-av1.md)
- [VAAPI](/quest/m1/video-vaapi.md) - existing Linux codec expansion
- [VP8/VP9 in OBS](/quest/m3/obs-moq-video/vpx-obs.md) - the software decoder reaching the OBS source
Comment thread
coderabbitai[bot] marked this conversation as resolved.
Resolved
4 changes: 2 additions & 2 deletions rs/moq-c/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -33,8 +33,8 @@ moq-mux = { workspace = true, features = ["async"] }
moq-net = { workspace = true }
moq-tokio = { workspace = true, features = ["quic", "websocket", "tcp", "uds"] }
# Select the NVIDIA and VAAPI hardware codecs explicitly because workspace
# defaults are off. The hardware backends dlopen their driver; `vaapi` also runs
# bindgen, so Linux builds need libclang. The software fallback is `openh264`.
# defaults are off. The hardware backends dlopen their driver, so neither needs
# anything on the build host. The software fallback is `openh264`.
moq-video = { workspace = true, features = ["nvidia", "vaapi"] }
# Blocking on `encode::Sink` from the synchronous C entry points; see
# `video::block_on` for why not a tokio helper.
Expand Down
10 changes: 5 additions & 5 deletions rs/moq-cli/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ path = "src/main.rs"
doc = false

[features]
default = ["cluster-lan", "quic", "websocket", "nvidia", "s3", "gcs", "azure"]
default = ["cluster-lan", "quic", "websocket", "nvidia", "vaapi", "s3", "gcs", "azure"]
# LAN discovery and meshing (`--cluster-lan`).
cluster-lan = ["moq-relay/cluster-lan"]
jemalloc = ["moq-tokio/jemalloc"]
Expand Down Expand Up @@ -70,10 +70,10 @@ nvidia = ["moq-video?/nvidia", "moq-transcode?/nvidia"]
# feature it turns on. Only bites when `capture`, `transcode`, or `play` pulls in
# moq-video.
openh264 = ["moq-video?/openh264", "moq-transcode?/openh264"]
# Intel/AMD VAAPI hardware encoding (Linux), opt-in like the moq-video feature
# it turns on: libva is dlopen'd at runtime, but moq-vaapi's bindgen needs
# libclang on the build host. Only bites when `capture`, `transcode`, or `play`
# pulls in moq-video.
# Intel/AMD VAAPI hardware codecs (Linux), on by default like the moq-video
# feature it turns on: libva is dlopen'd at runtime and moq-vaapi checks its
# bindings in, so nothing is needed on the build host. Only bites when
# `capture`, `transcode`, or `play` pulls in moq-video.
vaapi = ["moq-video?/vaapi", "moq-transcode?/vaapi"]
# The V4L2 M2M codecs of an ARM SoC (a Raspberry Pi's VideoCore and the like),
# opt-in like the moq-video feature it turns on. Costs the build nothing: moq-v4l
Expand Down
3 changes: 1 addition & 2 deletions rs/moq-ffi/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -52,8 +52,7 @@ moq-tokio = { workspace = true, default-features = true }
# cudarc / moq-nvenc / libloading / moq-vaapi sit behind
# `cfg(target_os = "linux")` and Android is `target_os = "android"`. On Linux
# both dlopen the driver, so the wrappers still link GPU-less and start
# driverless; `vaapi` alone costs the build something, since moq-vaapi's bindgen
# needs libclang (the dev shell carries it).
# driverless, and neither needs anything on the build host.
moq-video = { workspace = true, optional = true, features = ["nvidia", "vaapi"] }
# Blocking on `encode::Sink` from the synchronous producer exports; see
# `video::block_on` for why not a tokio helper.
Expand Down
18 changes: 9 additions & 9 deletions rs/moq-transcode/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -13,20 +13,20 @@ keywords = ["quic", "http3", "webtransport", "media", "video"]
categories = ["multimedia", "multimedia::video", "multimedia::encoding"]

[features]
# NVIDIA is on by default, mirroring moq-video: NVDEC -> NVENC is the zero-copy
# GPU pipeline, and it dlopens its driver so it needs nothing on the build host.
# OpenH264, the software H.264 fallback, is vendored C++ and opt-in like
# moq-video's: release builds enable it. `vaapi` stays opt-in because its
# bindgen needs libclang; `v4l2` mirrors moq-video's opt-in too, though it costs
# the build nothing.
default = ["nvidia"]
# NVIDIA and VAAPI are on by default, mirroring moq-video: each is a zero-copy
# GPU pipeline (NVDEC -> NVENC, or VA decode -> VPP -> VA encode), and each
# dlopens its driver, so neither needs anything on the build host. OpenH264, the
# software H.264 fallback, is vendored C++ and opt-in like moq-video's: release
# builds enable it. `v4l2` mirrors moq-video's opt-in too, though it costs the
# build nothing.
default = ["nvidia", "vaapi"]
# NVENC + NVDEC (Linux, dlopen'd at runtime; a no-op elsewhere).
nvidia = ["moq-video/nvidia"]
# Software H.264 encode/decode through moq-video's vendored OpenH264 backend.
# Off by default; enable it in release builds.
openh264 = ["moq-video/openh264"]
# Intel/AMD VAAPI hardware encoder (Linux; libva dlopen'd at runtime, but
# moq-vaapi's bindgen needs libclang at build time).
# Intel/AMD VAAPI hardware codecs (Linux; libva dlopen'd at runtime, and
# moq-vaapi checks its bindings in, so nothing is needed on the build host).
vaapi = ["moq-video/vaapi"]
# The V4L2 M2M codecs of an ARM SoC (Linux; moq-v4l checks its bindings in, so
# nothing is needed on the build host).
Expand Down
Loading
Loading