feat(client-linux): VAAPI hardware decode — zero-copy dmabuf into GraphicsOffload
ci / docs-site (push) Failing after 45s
ci / web (push) Failing after 32s
apple / swift (push) Successful in 1m16s
ci / rust (push) Failing after 1m18s
docker / build-push (., web/Dockerfile, punktfunk-web) (push) Successful in 5s
docker / build-push (ci, ci/fedora-rpm.Dockerfile, punktfunk-fedora-rpm) (push) Failing after 7s
docker / build-push (ci, ci/rust-ci.Dockerfile, punktfunk-rust-ci) (push) Successful in 7s
docker / build-push (docs-site, docs-site/Dockerfile, punktfunk-docs) (push) Successful in 5s
docker / deploy-docs (push) Has been skipped
deb / build-publish (push) Failing after 1m38s
rpm / build-publish (push) Successful in 4m10s

Stage 1.5: on Intel/AMD clients libavcodec's VAAPI hwaccel decodes on
the GPU; frames map to DRM-PRIME dmabufs (av_hwframe_map, zero copy)
and reach GTK as GdkDmabufTexture (BT.709 limited CICP color state —
GDK's dmabuf default is BT.601). Inside GtkGraphicsOffload that is the
decoder-to-subsurface path, direct-scanout eligible when fullscreen.

Fallback ladder, live-verified on the NVIDIA dev box: no VAAPI device
-> software decode at session start (logged reason); a mid-session
VAAPI error (e.g. broken nvidia-vaapi-driver) demotes to software and
the host's IDR/RFI recovery resynchronizes; a rejected dmabuf import
logs and the stream continues. PUNKTFUNK_DECODER=software|vaapi
overrides; the first-frame log now names the active path.

The hwaccel path is raw ffmpeg-sys FFI (ffmpeg-next wraps none of it):
hw device ctx + get_format pinned to AV_PIX_FMT_VAAPI (NONE on
mismatch so cpu-fallback never silently engages inside libavcodec),
thread_count=1, LOW_DELAY. Surface lifetime rides DrmFrameGuard into
the texture's release func — GDK runs it on both success and failure.

Needs an Intel/AMD client box (Steam Deck/Bazzite) to live-verify the
hardware path; the software path is unchanged and revalidated.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-06-12 23:26:59 +00:00
parent b5c30dff4f
commit 4b1bbfdf0e
4 changed files with 350 additions and 41 deletions
+49 -9
View File
@@ -212,20 +212,60 @@ pub fn new(
// --- Frame consumer: newest texture wins, set on the GTK frame clock's cadence. ---
{
let picture = picture.downgrade();
// The host encodes BT.709 limited-range; without an explicit color state GDK
// would convert NV12 dmabufs with the (BT.601) dmabuf default.
let rec709 = {
let cicp = gdk::CicpParams::new();
cicp.set_color_primaries(1);
cicp.set_transfer_function(1);
cicp.set_matrix_coefficients(1);
cicp.set_range(gdk::CicpRange::Narrow);
cicp.build_color_state().ok()
};
glib::spawn_future_local(async move {
while let Ok(f) = frames.recv().await {
let Some(picture) = picture.upgrade() else {
break;
};
let bytes = glib::Bytes::from_owned(f.rgba);
let tex = gdk::MemoryTexture::new(
f.width as i32,
f.height as i32,
gdk::MemoryFormat::R8g8b8a8,
&bytes,
f.stride,
);
picture.set_paintable(Some(&tex));
match f {
DecodedFrame::Cpu(c) => {
let bytes = glib::Bytes::from_owned(c.rgba);
let tex = gdk::MemoryTexture::new(
c.width as i32,
c.height as i32,
gdk::MemoryFormat::R8g8b8a8,
&bytes,
c.stride,
);
picture.set_paintable(Some(&tex));
}
DecodedFrame::Dmabuf(d) => {
let mut b = gdk::DmabufTextureBuilder::new()
.set_display(&picture.display())
.set_width(d.width)
.set_height(d.height)
.set_fourcc(d.fourcc)
.set_modifier(d.modifier)
.set_n_planes(d.planes.len() as u32)
.set_color_state(rec709.as_ref());
for (i, p) in d.planes.iter().enumerate() {
b = unsafe { b.set_fd(i as u32, p.fd) }
.set_offset(i as u32, p.offset)
.set_stride(i as u32, p.stride);
}
let guard = d.guard;
// GDK runs the release func whether the import succeeds or not.
match unsafe { b.build_with_release_func(move || drop(guard)) } {
Ok(tex) => picture.set_paintable(Some(&tex)),
Err(e) => {
// Import rejected (format/modifier) — surfaces once per
// session in practice; the stream continues on the next
// frame, and PUNKTFUNK_DECODER=software is the escape.
tracing::warn!(error = %e, "dmabuf texture import failed");
}
}
}
}
}
});
}