f57fc85d3468f1627ec7f014e9a150efe01a77f7
11
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  enricobuehlerandClaude Opus 5
|
306f4a514d |
refactor(pf-capture): structural splits + collapse the restated signal set (Phase 5)
Refactors LAST, after every defect fix, so no behaviour change hides in a move diff — and so the newly-exposed review surface was already clean when it landed (the WP7 discipline). 28f8fc71's recorded trap (an inline `use super::X` inside a moved body silently changing meaning) was audited first: every `use super::` in the moved code is either at module level — where `super` still means `linux` — or inside a `mod tests` that moved with its parent. 5.1 `mod pipewire` → `linux/pipewire.rs`. It was 1,900 of that file's lines. `mod.rs` is a directory module, so a plain `mod pipewire;` resolves with no `#[path]`. 5.2 out of that, `linux/pw_pods.rs` (the 7 param-pod builders + `serialize_pod` + the PQ constant and its test) and `linux/pw_cursor.rs` (`CursorState`, `decode_bitmap_pixel`, `update_cursor_meta`, `dst_offsets`, both `composite_cursor*`). The pods are the crate's WIRE surface — a missing property is not a compile error but a link that stalls in `negotiating`; the cursor half is producer-driven and bounds-critical. Both are now pure enough to unit-test without a compositor, which is what Phase 6 needs. 5.3 `linux/portal.rs`: the ScreenCast/RemoteDesktop handshake, the cursor-mode choice and GNOME's BT.2100 probe — the async/tokio/zbus control plane, separated from the realtime half. Zero re-exports changed: `mod.rs` re-exports `gnome_hdr_monitor_active` at its old path. 5.4 `idd_push/open.rs` (the whole one-time construction path + `SharedObjectSa` + `AttachTexFail`) and `idd_push/compose_kick.rs`. Read `open.rs` when a session will not START and the parent when one stops flowing; the steady state stays with the parent by decision. Also moved the ~65-line stall REPORTING block out of `try_consume` (the hot loop) into `StallWatch::report`, taking its two correlation counters with it — they were capturer fields nothing else touched. 5.5 `windows/dxgi/selftest.rs`: `p010_reference`, both self-tests, `hdr_p010_convert_bars_on_luid`, `f32_to_f16` and the two test modules — the validation path, none of which runs in a session. The two `pub` entry points are re-exported, so `main.rs`'s subcommand and pf-encode's live e2e keep their paths. (An explicit `#[path]`, like `idd_push`'s children: this file is itself reached through a `#[path]`, so a bare `mod` would resolve to `windows/selftest.rs`.) 5.6 `CaptureSignals`. The seven shared flags/slots were created in `spawn_pipewire`, `_cb`-cloned one by one, passed as seven of `pipewire_thread`'s parameters, and then re-declared as fields on `PwHandles`, `PortalCapturer` AND `UserData` — the same list written four times, each with its own drifting copy of the doc comment. One `#[derive(Clone)]` struct instead (a refcount bump per field), which also makes it structurally obvious that producer and consumer share ONE set. And `CaptureOpts` for the four trailing `bool`s: four adjacent same-typed positional arguments are a silent-transposition footgun — swap `want_444` and `want_hdr` and it compiles, negotiates the wrong pod family, and surfaces as a black screen ten seconds later. 5.7 trait shape, targeted: `set_active(&self)` → `&mut self` (it took `&self` only because the flag happened to be an `Arc<AtomicBool>` — an implementation detail leaking into the contract); `capture_target_id` ⇒ `resize_output`'s pairing rule stated on both methods and in a cluster note; one-line cluster headers over the 13 methods. NOT done from 5.7: taking the gamescope targets as an `open_*` option instead of the trait method. It would put a Linux-only parameter on the cross-platform `open_virtual_output` and on the host's `capture_virtual_output` facade — a `#[cfg]`-shaped wart in two shared signatures to delete one already-`cfg`'d defaulted method whose lifecycle rule 2.5c had already made harmless. Wrong trade under 5.7's own "no churn for no defect fixed" principle. File sizes: `linux/mod.rs` 2,778 → 770 (target ≤900 ✓), `windows/dxgi.rs` 1,374 → 954, and `windows/idd_push.rs` 2,694 → 1,922. The last two miss the plan's ≤850/≤1,300 targets, which were computed against the ORIGINAL line counts — Phases 1–4 added several hundred lines of SAFETY proofs and defect rationale to exactly these files before the split. The moves themselves are the ones the plan specifies; closing the rest would mean inventing new splits it does not call for. Zero behaviour delta. pf-capture 20/20; workspace clippy --all-targets clean on Linux and on windows-msvc. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
enricobuehlerandClaude Opus 5
|
530909a154 |
perf(pf-capture): take three per-frame costs off the Windows HDR hot path (Phase 4)
**4.1 (W10) — the HDR convert allocated two views and mapped a constant buffer per frame, inside
the ring slot's keyed-mutex hold.** So the driver's publisher sat blocked on that slot while the
host created `CreateRenderTargetView`s and Mapped/Unmapped a 16-byte buffer whose contents never
change. Both are lifetime-of-mode facts, not per-frame ones:
- the two P010 plane RTVs now belong to the out-ring SLOT, built once in `ensure_out_ring`
(`out_ring` becomes `Vec<OutSlot>`). A driver that rejects planar RTVs — the one hard
requirement of this path — therefore fails at ring build, where such a failure belongs, instead
of on the first frame.
- `inv_src` becomes an IMMUTABLE constant buffer filled in `HdrP010Converter::new(device, w, h)`.
The converter is already dropped by `recreate_ring` on every mode change, so it cannot go
stale. This also deletes the `Map`'s silently-ignored failure: it could leave the chroma pass
sampling at a garbage texel size with no error anywhere.
**4.2 (W15) — `sdr_white_level_scale` ran inside the keyed-mutex hold too.** It is a CCD query that
contends the display-config lock — precisely the contention `DescriptorPoller`'s doc says must stay
off the frame path — and `prepare_blend_scratch` called it while holding the slot. Moved to
`refresh_sdr_white_scale`, called at open and after each ring recreate (where the scratch is
invalidated anyway). "Only on scratch rebuilds" is not the same as harmless when the thing being
blocked is the producer.
**4.3 (W11) — `VideoProcessorSetStreamAutoProcessingMode(vp, 0, false)`.** The comment claimed "no
interpolation/auto-processing" while only setting the frame format; auto-processing's documented
default is ENABLED, so vendor denoise/edge-enhancement was free to run inside every SDR
`VideoProcessorBlt` — altering the pixels we encode and spending video-engine time on the
desktop-capture hot path. Now actually disabled, with each call's purpose stated separately.
**4.4 (the Linux CPU-path map cache) is deliberately NOT done.** The plan gates it on measurement
("measure first, and only if the CPU path still matters") and this box cannot measure a live capture
session; it also needs a frame-buffer return path through `FramePayload::Cpu`, which is a larger
change than the rest of this phase. Left for Phase 7's measured pass.
Compile-verified for windows-msvc locally; pf-capture 20/20 on Linux; workspace clippy
--all-targets clean on both targets. The GPU-capture check and the measured `cap_us` delta are owed
to Phase 7.3.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
enricobuehlerandClaude Opus 5
|
bc45287dc6 |
fix(pf-capture): Windows IDD-push defects — HDR pin, cursor, recreate, handles, f16 (Phase 3)
Compile-verified for x86_64-pc-windows-msvc locally (a scratch workspace symlinking the real
sources against a stub punktfunk-core — the `quic` feature's ring/opus C builds are what blocks an
in-tree cross-check). Behaviour is owed the on-glass validation in sweep Phase 7.3.
**3.1 (W1/F3) — the HDR pin asserted a flip it never verified.** `poll_display_hdr` discarded
`set_advanced_color`'s `bool` and then wrote `now.hdr = self.client_10bit` — the DESIRED state in
place of the observed one. On a display that cannot be flipped (the state this file already logs as
"Downgrade point D" at open) that broke in both directions: wanting HDR, the fabricated `true`
differed from `current`, so two poller samples drove `recreate_ring(true, …)` and rebuilt the ring
FP16 while the driver composed 8-bit BGRA — every publish dropped by the driver's format guard,
`recovering_since` expiring, `try_consume` bailing: a permanent 3-second reconnect loop. Wanting
SDR, the fabricated `false` MATCHED `current`, so no recreate ever fired and the ring stayed BGRA
against an FP16 composition — the same dropped-publish outcome, silently. Now it re-reads
`advanced_color_enabled` and follows what the display actually composes, with a one-shot error
naming want/observed/returned. A not-yet-settled read costs one debounce cycle, never a wrong ring,
which is why this does not block the frame path on a settle poll the way `open_on` does. Downgrade
point D's error now carries the same pair, so `Some(false)` (display says no) and `None` (the CCD
read failed) are distinguishable.
**3.2 (W2, W3, W6) — cursor correctness.**
- W2: the poller's desktop rect was captured once at open and used forever, for BOTH the
desktop→frame offset and the `in_rect` visibility test — while both mid-session mode-change
paths (`resize_output`, `poll_display_hdr` → `recreate_ring`) keep the same poller. After an
in-place resize the pointer was clipped to the old rect and offset by a stale origin. It is now
a SEED: the poll thread re-queries on its existing 250 ms reattach cadence, keeping the last
good value on `None` (a transient CCD failure must not park the rect at zero and report every
position invisible), which keeps the CCD call off the encode thread as `DescriptorPoller`
demands.
- W3: `composite_forced` tested `cursor_sender.is_none()`, but §8.6's rationale is "no cursor
CHANNEL" — and the delivery just above it is explicitly allowed to fail non-fatally, which is
precisely the state needing the rescue. It was the one state that skipped it: a negotiated
channel that failed to create or deliver left a cursor-excluded target with NO pointer at all.
Now `cursor_shared.is_none()`, evaluated after that binding.
- W6: `cursor()` degraded poller→shm correctly, but the BLEND path — the only consumer that
matters in the composite model, since the Windows encode loop never attaches `frame.cursor` —
read the poller directly with no `alive()` check and no fallback, so the documented fallback and
the spawn-failure warning were both untrue for exactly those sessions (a dead poller meant
pointer-less frames, not a degraded pointer). One `live_cursor()` now serves all three
consumers and LATCHES the source, because the two keep independent serial namespaces and
interleaving them poisons the client's shape cache.
**3.3 (W4, W5, W14) — recreate hardening.**
- W4: `recreate_ring` committed `display_hdr`/`width`/`height` BEFORE the fallible
`create_ring_slots` (VRAM pressure at a large new mode — exactly when resizes happen), leaving
a failed recreate emitting frames stamped with the new geometry against the old ring, the old
generation and an unchanged header. Slots are built first; nothing after the commit point fails.
- W5: a recreate never cleared the driver's status words, and `wait_for_attach` — the only
classifier of TEX_FAIL/BIND_FAIL and the only source of the LUID rebind — runs at open ONLY. A
stale `OPENED` therefore made a failed re-attach look healthy while the recover-or-drop bail
reported nothing. Now cleared before the Release generation store (plus `status_logged`), and
the 3 s bail prints the live `(driver_status, detail, render_luid)` the way `next_frame`'s 20 s
bail already did. The four-field read is one `driver_diag()` helper instead of four copies of
the same unsafe block.
- W14: `IDD_GENERATION` is a full `u32` but the publish token carries 24 bits and `unpack` masks
what it reads, so past 2²⁴ recreates `tok.generation != self.generation` would be permanently
true — every frame rejected. Masked at the single mint point, and 0 skipped (it is also the
cleared-`latest` sentinel).
**3.4 (W8, W9, W12) — handle hygiene.** `shared_object_sa`'s security descriptor is a `LocalAlloc`
nobody freed: leaked twice per open and once per ring recreate. It is now an RAII `SharedObjectSa`
whose `Drop` `LocalFree`s it and whose `as_ptr()` only lends a borrow — which also makes the
"descriptor must outlive the attributes" rule structural instead of a comment. The PyroWave fence's
shared NT handle, created per capturer and never closed, becomes an `OwnedHandle` (the encoder holds
its own duplicate, so closing ours is safe). `cursor_blend`'s `cbuf_scale` is cached only on a
successful `Map` — caching unconditionally wedged the HDR/SDR cursor scale for the session after one
transient failure.
**3.5 (W7) — `f32_to_f16` swallowed the rounding carry.** `sign | half_exp | (half_mant + round)`
ORs a mantissa carry into bit 10, so for every ODD biased exponent (bit 10 already set) the carry
vanished and the result came back ~2× low: `1.9998779 → 1.0`, `0.49996948 → 0.25`. Only values one
ULP below a power of two are affected — precisely what a gradient test pattern is full of — so this
made `hdr-p010-selftest` FAIL a correct shader. Composed additively, with 4 tests (18 asserted bit
patterns, a round-trip property over the self-test's scRGB values, saturation) — all verified
numerically against a standalone reference on this box, including a scan confirming old-vs-new
diverges ONLY on the carry cases. Phase 0.1's `--all-targets` lint is what lets these compile in CI
at all.
pf-capture 20/20 on Linux; workspace clippy --all-targets clean on Linux and windows-msvc.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
enricobuehlerandClaude Opus 5
|
54a37aeb46 |
docs(pf-capture): truth pass over comments, docs and log strings (sweep Phase 1)
Every item here is prose that actively misdirects a maintainer or an operator. Landing it before
the defect fixes means those diffs get reviewed against accurate comments.
1.1 (X5) — split two merged doc blocks that documented the item ABOVE them: `hdr_meta`'s whole
contract was glued onto `fn cursor()` in lib.rs (leaving `hdr_meta` undocumented), and
`hdr_p010_selftest_at`'s was glued onto `hdr_p010_convert_bars_on_luid` in dxgi.rs. rustdoc for
the trait's central HDR contract was simply wrong.
1.2 (L9 doc half) — the `Capturer` trait doc advertised a "bounded drop-oldest channel". The
Linux portal's `sync_channel(8)` + `try_send` is drop-NEWEST: under a consumer stall the arriving
frame is discarded and the encoder gets the stalest queued one. Say so. (Phase 2.6 changes the
behaviour; this commit only stops the doc from lying about today's code.)
1.3 — delete the DONT_FIXATE claim from the dmabuf offer comment. `build_dmabuf_format` emits a
plain MANDATORY `ChoiceEnum::Enum` and `param_changed` re-emits nothing, so there is no two-step
DMA-BUF handshake here; libspa 0.9's `ChoiceFlags` cannot even express DONT_FIXATE. Recorded as a
real follow-up instead of an implemented thing.
1.4 — `mainloop.run()` no longer "blocks until process exit": the quit channel attached above it
stops the loop from `PortalCapturer::drop`.
1.5 (L11) — the 6-arm negotiation log ladder told a fully zero-copy PyroWave session "VAAPI
encode with the CPU capture path — zero-copy was disabled", one line after logging that it had
advertised the PyroWave device's dmabuf modifiers: the passthrough arm was gated on
`pyrowave_modifiers.is_empty()`, so the pyrowave case fell through to the CPU warning. Ungate it;
the CPU-path arm is now reachable only when no dmabuf is advertised at all, which is what it
claims. Its parenthetical also gains the third real cause (the session asked for CPU frames).
Control-flow-neutral for every other combination.
1.6 (W13) — `kick_dwm_compose` claimed a "sub-millisecond" round trip while its
cursor-on-a-sibling-display branch sleeps 35 ms on the CALLER's thread. State the cost on the
function and at both call sites, including which one is on the live frame path.
1.7 (W16) — DDA-era residue, DDA having been removed:
- retarget the win32u hook's rationale + all four log strings: its job is no longer "keep DDA on
one adapter" but "keep the virtual display on the adapter SET_RENDER_ADAPTER pinned" (a DXGI
reparent surfaces as the driver's TEX_FAIL render-adapter mismatch). The hook stays installed.
- the per-monitor-v2 DPI awareness rationale likewise: not DuplicateOutput1's E_ACCESSDENIED any
more, but keeping cursor/window coordinates in the PHYSICAL pixels the host's CCD geometry
(`source_desktop_rect`, `desktop_bounds`) is already in.
- `HYBRID_HOOK_HITS` was write-only. Surface it as `hybrid_hook_hits()` on the IDD-push open
line, which is the first point where DXGI has actually been exercised — the patch-readback
check proves the bytes landed, only this proves DXGI reaches the export.
- delete `hdr_p010_selftest()`: an unreachable 64×64 wrapper (main.rs calls
`hdr_p010_selftest_at`); its description moves onto the function that survives.
- `VideoConverter`'s docs promised a P010/BT.2020 output and a live scRGB input path. It pins
`YCBCR_STUDIO_G22_LEFT_P709` unconditionally and its only caller always passes
`scrgb_input: false`.
1.8 (X7) — the native handshake's 4:4:4 gate comment stated the OPPOSITE of what
`pf_capture::capturer_supports_444` returns on Windows ("delivers subsampled NV12/P010 today, so
it returns false there" — it forwards `resolved_backend_ingests_rgb_444()` and the IDD ring passes
BGRA through for an SDR 4:4:4 session).
No `.rs` behaviour delta: comments, doc comments and log strings, plus 1.5's control-flow-neutral
ladder and 1.7's counter accessor. Linux tests 6/6; clippy --all-targets clean on both targets.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
enricobuehlerandClaude Opus 5
|
6a2a153c0b |
chore(pf-capture): make the crate verifiable — Windows CI lint + two denies (sweep Phase 0)
Gate for the rest of design/pf-capture-sweep.md: today half this crate is compiled by no CI job
at all, so the Windows-side findings can't even be type-checked.
0.1 (X1) — windows-host.yml lints `-p pf-capture --all-targets`. The host lint above it builds
pf-capture only as a DEPENDENCY, so its `#[cfg(test)]` modules were never compiled anywhere: the
5 StallWatch tests, the DXGI HDR self-tests and the cursor-conversion tables were dead code in
CI. The workflow's own comment already diagnosed this blind spot and fixed it for pf-encode;
pf-capture never got the same treatment. No features on this crate, so no feature juggling and
no extra dep tree.
0.2 (X2) — `#![deny(unsafe_op_in_unsafe_fn)]` beside the existing
`deny(clippy::undocumented_unsafe_blocks)`. The crate declares "every unsafe block carries a
SAFETY proof; enforce it" in ten file headers, but in edition 2021 that lint has nothing to fire
on inside an `unsafe fn` — so the hardest FFI in the crate was exempt from its own program: the
ring/slot construction, the fence signal, the blend scratch, and every D3D converter ctor/convert.
119 operations across 16 functions now carry a proof. `shared_object_sa` loses its `unsafe`
marker instead (it states no precondition — only its RESULT is a raw pointer, which is the
caller's business).
0.3 (X4) — drop the crate-wide `#![allow(dead_code)]`. Verified: zero warnings on either target
without it, so it was hiding nothing the lint can see (W16's dead items are `pub`/written-once,
so they need Phase 1.7's deletion instead).
No behaviour change: only lint attributes, `unsafe { }` scoping and comments.
Linux `cargo test -p pf-capture` 6/6, clippy --all-targets clean on both targets (Windows
verified by cross-checking x86_64-pc-windows-msvc locally).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
||
|
enricobuehlerandClaude Opus 4.8
|
0af280a793 |
feat(windows/capture): host-side cursor compositing for the capture model — the deterministic fix
Root cause, proven on-glass across five driver builds: a declared IddCx hardware cursor is IRREVOCABLE. There is no un-declare DDI, the empty-caps re-setup is rejected INVALID_PARAMETER, and after a successful same-mode re-commit with the driver's re-declare provably suppressed (sticky per-target flag, zero re-setups logged) DWM still never composites the software cursor back into the monitor's frames. Every DWM-based composite flip is therefore unfixable. The composite (capture) model is now implemented in OUR pipeline: - the driver keeps its hardware cursor declared for the session's whole life (the state that works) — frames stay pointer-free always; - in composite mode try_consume routes the conversion through a blend scratch: slot copy + one alpha-blended quad of the GDI poller's shape at its polled position (CursorBlendPass — fullscreen-triangle VS with viewport placement, straight-alpha PS, sRGB→linear for FP16/HDR rings), covering all four convert paths (NV12, 4:4:4 copy, P010, PyroWave) GPU-side under the slot's keyed mutex; - pointer-only motion produces no driver publish (declared hw cursor), so try_consume REGENERATES from the last slot whenever the polled cursor state changes — the cursor moves on a static desktop at tick rate, without faking freshness for the stall/death watchdogs; - set_cursor_forward becomes purely host-internal state; the SET_CURSOR_FORWARD IOCTL machinery is no longer used by the host (driver keeps it, dormant) and its sender plumbing is removed. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> |
||
|
enricobuehlerandClaude Fable 5
|
428bd2519c |
test(qsv): converter→ring-profile-P010→encoder live e2e (the RTV-written seam)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
enricobuehlerandClaude Fable 5
|
87d7eceabf |
fix(encode/qsv): write the colour VUI unconditionally + Intel-lane colour diagnostics
The native QSV encoder only attached mfxExtVideoSignalInfo on HDR sessions — an SDR stream carried no colour description at all (NVENC writes 709-limited unconditionally; qsv.rs now mirrors it). Diagnostics for the field-reported blue/magenta Intel-host colours: - hdr-p010-selftest takes an optional WxH + GPU vendor (intel|nvidia|amd) and prints which adapter it tested — 64x64 on the default GPU proved nothing on dual-GPU boxes, and the field heights (1080/1400) are not 16-aligned. - pf-capture: ignored live test running the converter at 1920x1080 pinned to the Intel adapter. - pf-encode: qsv_live_p010_1080_colorbars_dump — known P010 bars through the UNALIGNED-height ingest copy (1080 src -> align16 1088 pool, the seam no 640x480 test exercises) to Main10 HEVC, dumped for off-box decode checks. - dxgi.rs: drop the stale 'falls back to the R10 path' claim (no such path). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
enricobuehlerandClaude Fable 5
|
188edde2b3 |
feat(pyrowave): Windows host HDR + 4:4:4, Rust client HDR present
Phase 3 of design/pyrowave-444-hdr.md. A PyroWave session now negotiates HDR
(10-bit) and 4:4:4 on a Windows host exactly like HEVC/AV1, and the Linux
client presents it through the real HDR10 path.
Host (Windows): BgraToYuvPlanes becomes mode-aware — SDR/BGRA and HDR/scRGB
variants at half- or full-res chroma. The HDR passes reuse HdrP010Converter's
exact colour math (scRGB -> PQ BT.2020 limited studio codes, verified by
hdr_p010_selftest) but write P010-style MSB-packed codes into two separate
shareable R16_UNORM/R16G16_UNORM textures; chroma keeps the pyrowave family's
centre-sited 2x2 box. idd_push pins the composition to the NEGOTIATED depth
(SDR sessions force advanced color off as before; 10-bit sessions enable it
and ride the FP16 ring), and the descriptor poller re-asserts that state
instead of following display flips the fixed-format encoder can't. The
encoder imports 8/16-bit planes per session and stamps the sequence header's
BT.2020/PQ/matrix bits on HDR (stamp_color_bits, extending 574e3e4e's range
stamp); supports_10bit/can_encode_10bit/can_encode_444 gates open (HDR
Windows-only — Linux capture has no HDR source).
Client: the plane ring becomes R16_UNORM for 10-bit sessions (with a
STORAGE_IMAGE format probe), the planar CSC pass joins the HDR10 swapchain
rebuild (set_hdr_mode previously destroyed it without rebuilding — latent),
st.hdr follows frame.color.is_pq(), and the planar push constants carry
depth-10 MSB-packed rows + the PQ tonemap mode, identical to the NV12 arm.
Verified: .173 (RTX 4090) deploy-config clippy + fmt + wire tests + the
extended pyrowave_win_smoke (10-case {SDR,HDR}x{420,444} matrix incl. R16
imports and header stamps); .21 (RTX 5070 Ti) clippy across 4 crates, host
186 tests, client/presenter/encode tests, both Linux GPU smokes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
||
|
enricobuehlerandClaude Opus 4.8
|
ebd9967547 |
feat(pyrowave): Windows host encoder — separate-plane zero-copy D3D11→Vulkan
Wire PyroWave into the Windows host (design/pyrowave-windows-host-zerocopy.md). Before this a macOS client + Windows host that both selected PyroWave silently ran HEVC: the host never advertised CODEC_PYROWAVE and open_video_backend bailed. Approach (zero-copy, no GPU→CPU→GPU): pyrowave owns its own Vulkan device (create_device_by_compat, by render-GPU vendor/device-id — NOT LUID, invalid in Session 0). The capturer runs a BGRA→YUV BT.709-limited CSC (matching rgb2yuv.comp) into TWO SEPARATE shareable plane textures — full-res R8 Y + half-res R8G8 CbCr — which the encoder imports into pyrowave's device. Separate single/two-component textures import reliably on NVIDIA at any size; a single planar NV12 import does NOT (the vendored interop test: "only very specific resource sizes" — confirmed on-glass: 1024² fine, 720p/1080p/1440p garbage). A shared D3D11 fence, signalled after the CSC, is imported as a Vulkan timeline semaphore so the wavelet read is ordered after it. - pf-encode: enc/windows/pyrowave.rs (Encoder impl, two-plane import + Linux-style plane views); host_wire_caps advertises CODEC_PYROWAVE on Windows when the backend isn't Software; open_video_backend routes a negotiated PyroWave session first; pyrowave-sys on the Windows target; interop confirmed at open → clean HEVC fallback. - pf-encode: shared, unit-tested enc/pyrowave_wire.rs (single source of truth for the client-facing AU framing); Linux encoder uses it too. - pf-capture: dxgi.rs BgraToYuvPlanes CSC; idd_push.rs pyrowave mode — forces the virtual display SDR (the VideoProcessor can't ingest the FP16 HDR ring), a two-plane shareable out-ring, a shared fence passed every frame (so a rebuilt encoder re-imports it). Threaded via OutputFormat::pyrowave. - pf-frame: D3d11Frame::pyro carries the CbCr plane + fence; OutputFormat::pyrowave. Verified on .173 (RTX 4090): full-host build + clippy -D warnings (nvenc,amf-qsv) + fmt --all --check; pyrowave_wire unit tests; pyrowave_win_smoke GPU test round-trips distinct Y/Cb/Cr (100/180/60) exactly at 1024²/720p/1080p/1440p; Stage-0 interop validated in the real Session-0 service context on-glass. Deployed to the box. Owed: final on-glass picture/latency confirmation. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> |
||
|
enricobuehlerandClaude Opus 4.8
|
94ca4041ca |
refactor(host/W6.2): extract the frame-capture backends into the pf-capture crate
capture/linux (PipeWire portal) + capture/windows (IDD direct-push: dxgi mechanics, idd_push + submodules, synthetic_nv12) + pwinit move into crates/pf-capture behind the Capturer trait + synthetic sources (plan §W6). The crate speaks pf-frame (CapturedFrame/PixelFormat + the DXGI identity), pf-zerocopy (CUDA import), and the pf-win-display leaves, and NEVER pf-encode — the capture->encode edge is one-way. This completes the deliberate capture/encode crate split (the invasive path the plan had merged into one pf-media): capture and encode are now separate subsystem crates sharing only pf-frame. Four seams keep the capturer off the orchestrator: - VirtualOutput is EXPLODED into primitives (remote_fd/node_id/preferred_mode/ keepalive) by the host facade, so pf-capture never depends on the vdisplay type; - FrameChannelSender: the sealed-channel delivery is a Send+Sync closure the host facade builds from the pf-vdisplay control device + send_frame_channel IOCTL and hands in; ChannelBroker holds the closure instead of the control HANDLE (the whole-desktop handle-duplication security boundary is byte-for-byte unchanged); - console_session_mismatch + desktop_bounds live in pf-win-display (leaf peers); - pwinit moves here (audio caller -> pf_capture::pwinit). The host keeps capture.rs as a thin BRIDGE: it re-exports the vocabulary + capturer types (every crate::capture::* path is unchanged) and keeps open_portal_monitor / capture_virtual_output, which resolve the ZeroCopyPolicy + FrameChannelSender and call into pf-capture. verify_is_wudfhost + install_gpu_pref_hook are re-exported (the gamepad-channel bootstrap + the main.rs subcommand consume them). Co-developed: the resident-HID-mouse compose-kick hook (HID_COMPOSE_KICK + the HID-first cursor kick + _display_wake) rides this commit into pf-capture; the host mouse_windows registration side lands separately on top. Verified: Linux clippy -D warnings (pf-capture + host nvenc,vulkan-encode,pyrowave --all-targets) + host tests 299/299; Windows clippy -D warnings (pf-capture --all-targets + host nvenc,amf-qsv --all-targets) Finished exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> |