Files
punktfunk/crates/pf-ffvk/src/lib.rs
T
enricobuehler d6647b9183 feat(clients/windows): port the Vulkan session client to Windows — session-always
The punktfunk-session Vulkan client (clients/linux-session, now clients/session)
builds and runs on Windows; the WinUI shell spawns it for every stream. Verified
live: 10-bit HEVC via Vulkan Video on both AMD (iGPU) and NVIDIA, 5120x1440 at
130 fps / 8 ms end-to-end on the RTX 4090.

- pf-ffvk: Windows bindgen branch (FFMPEG_DIR + PF_FFVK_VULKAN_INCLUDE, no
  pkg-config); provisioning fetches Vulkan-Headers (pinned v1.4.309).
- pf-client-core: builds on Windows — WASAPI audio (audio_wasapi.rs, cfg-swapped
  via #[path], same surface as the PipeWire twin), VAAPI/dmabuf gated inline
  (chain = vulkan -> software), trust reads the WinUI shell's %APPDATA% stores
  (parity tests pin both serialized shapes), Settings gains adapter/hdr_enabled
  (serde-defaulted; Linux stores unaffected).
- pf-presenter: builds on Windows — dmabuf module Linux-gated; SDL keyboard grab
  while captured (Alt+Tab/Win reach the host); pick_device ranks discrete over
  integrated (device 0 was the iGPU on hybrid boxes — the silent footgun) and
  honors PUNKTFUNK_VK_ADAPTER (the Settings GPU pick, exported by the session).
- run loop: block in one SDL wait woken by input AND decoded frames (a per-
  session forwarder pushes a FrameWake user event) instead of a 1 ms poll —
  measured 111%% -> 5%% of a core (NVIDIA), 86%% -> 3.5%% (AMD), stats unchanged.
  The pump's decode-fence wait became once-per-window sampling (no per-frame
  pipeline stall; the stat now shows true backlog).
- pf-console-ui: builds on Windows (skia-safe msvc prebuilts); font lookup falls
  through fontconfig aliases to concrete DirectWrite families (Consolas/Segoe UI)
  — browse/coverflow works, verified against a live host.
- WinUI shell: session-always via new src/spawn.rs (GTK spawn.rs port —
  CREATE_NO_WINDOW, stdout contract, kill handle); the Stream screen is a status
  card (chips + stage lines from the child's stats). The legacy in-process
  D3D11VA path stays behind Settings "Streaming engine" / PUNKTFUNK_BUILTIN_
  STREAM=1 as the A/B baseline until Phase 8 deletes it. SessionParams.video_caps
  makes the HDR toggle real.
- clients/linux-session renamed to clients/session (builds for both OSes).
- CI/MSIX: both workflows build/test both bins with widened path filters; the
  MSIX ships punktfunk-session.exe. ARM64 session builds --no-default-features
  (rust-skia has no aarch64-pc-windows-msvc prebuilts; flip when it does).

A/B on this box (5120x1440 HEVC vs home-worker-5): NVIDIA Vulkan 130 fps / 8 ms
e2e / 1.6 ms decode — clearly better than the built-in path. The AMD iGPU VCN
saturates at ~52 fps where its own D3D11VA does ~70 — Adrenalin Vulkan decode is
slower on APU silicon; discrete RDNA validation gates Phase 8.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-08 23:21:36 +02:00

95 lines
3.4 KiB
Rust

//! FFmpeg's Vulkan hwcontext surface (`AVVulkanDeviceContext`, `AVVulkanFramesContext`,
//! `AVVkFrame`), bindgen-generated from the system headers at build time — see build.rs
//! for why this must not be hand-transcribed.
//!
//! The raw bindings use vulkan.h's own handle types (pointers on 64-bit). The [`ash`]
//! conversion helpers below cross between them and ash's u64-newtype handles; both sides
//! are the same underlying Vulkan object handles, so the casts are value-preserving.
#![allow(non_upper_case_globals)]
#![allow(non_camel_case_types)]
#![allow(non_snake_case)]
#![allow(clippy::missing_safety_doc)]
// bindgen's layout tests deref-null-pointer by design; silence the lints they trip.
#![allow(deref_nullptr)]
#![allow(unnecessary_transmutes)]
#[cfg(any(target_os = "linux", windows))]
include!(concat!(env!("OUT_DIR"), "/bindings.rs"));
/// Conversions between the generated vulkan.h handle types and ash's.
#[cfg(any(target_os = "linux", windows))]
pub mod ashx {
use super::*;
use ash::vk::Handle as _;
/// vulkan.h non-dispatchable handles are `*mut T` on 64-bit; ash's are `u64`
/// newtypes. Same bits either way.
pub fn image(h: VkImage) -> ash::vk::Image {
ash::vk::Image::from_raw(h as u64)
}
pub fn semaphore(h: VkSemaphore) -> ash::vk::Semaphore {
ash::vk::Semaphore::from_raw(h as u64)
}
// bindgen's enum repr is target-dependent: u32 on Linux (clang default), i32 on
// MSVC — so the cast is required on one target and a same-type no-op on the other.
#[allow(clippy::unnecessary_cast)]
pub fn image_layout(l: VkImageLayout) -> ash::vk::ImageLayout {
ash::vk::ImageLayout::from_raw(l as i32)
}
// --- ash → vulkan.h (filling AVVulkanDeviceContext) ---------------------------------
pub fn to_instance(h: ash::vk::Instance) -> VkInstance {
h.as_raw() as VkInstance
}
pub fn to_physical_device(h: ash::vk::PhysicalDevice) -> VkPhysicalDevice {
h.as_raw() as VkPhysicalDevice
}
pub fn to_device(h: ash::vk::Device) -> VkDevice {
h.as_raw() as VkDevice
}
/// ash's loader-level `vkGetInstanceProcAddr` as the header's PFN type. Both are the
/// same C ABI function pointer (`extern "system"` == `extern "C"` on the platforms
/// this crate builds for).
pub fn to_get_proc_addr(
f: unsafe extern "system" fn(
ash::vk::Instance,
*const std::ffi::c_char,
) -> ash::vk::PFN_vkVoidFunction,
) -> PFN_vkGetInstanceProcAddr {
unsafe { std::mem::transmute(f) }
}
}
#[cfg(all(test, any(target_os = "linux", windows)))]
mod tests {
use super::*;
/// The allocator runs (links against the system libavutil) and the struct is
/// readable at the offsets bindgen computed — sem_value zero-initialized.
#[test]
fn vk_frame_alloc_links_and_zeroes() {
unsafe {
let f = av_vk_frame_alloc();
assert!(!f.is_null(), "av_vk_frame_alloc returned NULL");
assert_eq!((*f).sem_value[0], 0);
assert_eq!((*f).queue_family[0], 0);
// Leak the one test frame rather than binding av_free here.
}
}
/// AV_NUM_DATA_POINTERS-sized arrays came through with the right length.
#[test]
fn frame_arrays_are_av_num_data_pointers() {
let f: AVVkFrame = unsafe { std::mem::zeroed() };
assert_eq!(f.img.len(), 8);
assert_eq!(f.sem_value.len(), 8);
}
}