enricobuehlerandClaude Opus 5
bc45287dc6
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>
Low-latency desktop and game streaming with first-class Linux and Windows hosts.
Run the host on a Linux machine or a Windows PC, connect from a Mac, PC, phone, tablet, or TV, and stream your desktop or games — each device at its own native resolution and refresh rate, over your local network.
📖 Documentation: docs.punktfunk.unom.io — start with How It Works or the Quick Start.
💬 Community: Discord — chat, support, and Android beta access · r/Punktfunk.
🔒 Security: found a vulnerability? Report it privately to security@punktfunk.com — see SECURITY.md. Please don't open a public issue.
punktfunk pairs a virtual-display streaming host with native clients on every platform. It speaks
the existing GameStream protocol, so any Moonlight client works
day one — and adds its own faster punktfunk/1 protocol that breaks the ~1 Gbps FEC wall with a
GF(2¹⁶) Leopard-RS transport. A single shared Rust core (punktfunk-core) holds the
protocol, FEC, and crypto, linked into the host and every native client — directly as a Rust crate
on Linux and Windows, and over a stable C ABI from the Apple and Android apps.
What makes it different
- Your device's exact mode. For each client that connects, the host spins up a virtual display sized to that device — 1080p60 to a laptop, 1440p120 to a desktop, 4K to a TV, all at once. No letterboxing, no scaling, no rearranging your real monitors.
- Displays you configure, not just create. Keep a game's display (and the game) alive across disconnects so a reconnect drops straight back in; make the stream your sole desktop or extend alongside your monitors; let several devices become monitors of one desktop; keep each client's scaling. One-click presets in the console — a dedicated couch box, a shared desktop, a multi-monitor workstation. See Virtual displays.
- A real virtual display on Windows, too. On Linux the host uses per-compositor virtual outputs; on Windows you get the same on-the-fly virtual display — at the client's exact mode, no physical monitor or dummy HDMI plug, even on the secure desktop (UAC / lock screen). It also has its own indirect display driver (IDD) the host pushes finished frames straight into, rather than scraping a screen — tight, push-based integration that's unusual for a Windows streaming host.
- Low latency, GPU end to end. Frames go straight from the compositor to the NVENC encoder with zero CPU copies (dmabuf → CUDA/Vulkan → NVENC), over a transport tuned for responsiveness rather than throughput. Stable 240 fps at 5120×1440; sub-millisecond capture-to-reassembly on-box, ~1.3 ms cross-machine on a LAN. (AMD/Intel encode via VAAPI, and a GPU-less software H.264 encoder exists as a fallback.)
- Works with what you already have. Any Moonlight/Artemis client connects over GameStream — and
native apps for macOS, Linux, Windows, and Android use the lower-latency
punktfunk/1protocol. - Secure by default. Hosts require a one-time SPAKE2 PIN pairing; after that, devices reconnect on a pinned identity. No accounts, no cloud. Hosts auto-advertise over mDNS, so clients find them on the network without typing an IP.
Status
| Component | State |
|---|---|
Core — punktfunk-core + C ABI (protocol · FEC · crypto · QUIC) |
✅ Complete & hardened |
| GameStream host → stock Moonlight | ✅ Live end-to-end: pairing, RTSP, audio, per-client virtual output at native resolution, GPU zero-copy NVENC, gamepads |
Native protocol — punktfunk/1 |
✅ Validated live: QUIC control + GF(2¹⁶) FEC/AES-GCM data plane, PIN pairing, mDNS discovery, mid-stream mode renegotiation |
| Windows host (Windows 11 22H2+, x64) | 🟡 Implemented & shipping as a signed installer: its own all-Rust IddCx virtual display (secure-desktop capable) with a sealed IDD-push capture path — finished frames pushed straight into its own driver, not screen-scraped (no DDA/WGC) · GPU encode (NVENC on NVIDIA, AMF/QSV on AMD/Intel, software H.264 without a GPU) · WASAPI audio · bundled virtual-gamepad drivers (no ViGEmBus) · HDR incl. Vulkan-game HDR. NVIDIA live-validated; AMD/Intel CI-green |
macOS / iOS / tvOS client (clients/apple) |
✅ Streaming live: VideoToolbox decode (HEVC, and AV1 on hardware that decodes it), controllers incl. DualSense, discovery, pairing, speed test |
Linux client (clients/linux + clients/session) |
✅ Streaming live: relm4/GTK4 launcher shell that spawns a Vulkan session binary — Vulkan Video / VAAPI / software decode, PipeWire audio, SDL3 controllers, Skia console UI; ships as Flatpak/apt/rpm/Arch |
Android client (clients/android, phone + TV) |
✅ Streaming live: AMediaCodec decode + HDR10, AAudio audio, controllers, discovery, pairing |
Windows client (clients/windows, WinUI 3) |
✅ Streaming live: WinUI 3 shell + Vulkan session presenter, hardware decode on all GPU vendors via Vulkan Video → D3D11VA → software (NVIDIA + Intel validated on glass), WASAPI audio, SDL3 controllers, discovery, pairing; ships as signed MSIX (x64 + ARM64). HDR10 implemented, on-glass validation pending |
Web console + management API (web/) |
✅ TanStack console over the OpenAPI mgmt API: host status, paired devices, on-demand PIN pairing, GPU selection, performance capture graphs, live host logs |
Every native client also ships a tiered stats overlay (Compact / Normal / Detailed) with a
shared vocabulary across platforms, and the session client carries a full gamepad-driven console
shell (pf-console-ui): host list, PIN pairing, settings, and an on-screen keyboard.
The GameStream host works with a stock Moonlight client — validated live on NVIDIA hardware
(RTX 5070 Ti, RTX 4090): PIN pairing that persists across restarts, an app catalog, RTSP/ENet/audio,
and video at the client's exact resolution and refresh via a per-session virtual output (KWin,
gamescope, Mutter, and Sway/wlroots backends), encoded with GPU zero-copy (dmabuf → CUDA/Vulkan →
NVENC) up to 5120×1440@240. The native punktfunk/1 protocol adds a QUIC control plane and a
GF(2¹⁶) Leopard-FEC + AES-GCM data plane (p50 ~0.8 ms capture→received at 720p120), with
mid-stream mode renegotiation and a wall-clock skew handshake so latency stays valid across machines.
Both run from one process: bare punktfunk-host serve is the secure native-only default
(punktfunk/1 + the management API/web console), and serve --gamestream additionally enables the
GameStream/Moonlight-compat planes (opt-in, trusted-LAN only — GameStream has inherent on-path
weaknesses). The host is managed through a REST API and web console. Builds against FFmpeg 7 or 8.
Full milestone status: docs.punktfunk.unom.io/docs/status · roadmap: /docs/roadmap.
Install the host
Pick your platform and install from its package registry — the per-platform guide covers adding the repo, first run, and the web console. The Linux host is the primary, most battle-tested path; a Windows host also ships as a signed installer (all-vendor: NVIDIA, AMD, Intel).
| Platform | Install | Guide |
|---|---|---|
| Ubuntu / Debian (apt) | sudo apt install punktfunk-host (after adding the repo) |
Ubuntu — GNOME · KDE |
| Bazzite / Fedora Atomic (systemd-sysext) | sudo bash punktfunk-sysext.sh install (no layering, no reboot; rpm-ostree + bootc also supported) |
Bazzite |
| Fedora (dnf) | dnf install punktfunk punktfunk-web (after adding the repo) |
Fedora — KDE |
| Arch / Steam Deck (pacman / sysext) | pacman -Sy punktfunk-host (binary repo) · sysext .raw (SteamOS) |
packaging/arch |
| Windows (11 22H2+, x64) | signed setup.exe from the package registry |
Windows Host |
punktfunk-host is the streaming host; punktfunk-web is the browser console (pairing + status).
After install, run punktfunk-host serve inside your desktop session (the secure native default;
add --gamestream on a trusted LAN if you also want stock Moonlight clients), then pair from the web
console. Full instructions: docs.punktfunk.unom.io/docs/install.
Connect a client
| Streaming to… | Use |
|---|---|
| Mac, iPhone, iPad, Apple TV | The Apple app (clients/apple) — also on TestFlight |
| Linux desktop / laptop, Steam Deck | punktfunk-client (Flatpak / apt / rpm / Arch) |
| Android phone or TV | The Android app (clients/android) |
| Windows | Native punktfunk-client (signed MSIX) or Moonlight |
| Anything else (browser, old phone, smart TV) | Moonlight over GameStream |
Each client discovers hosts on the network automatically and does a one-time PIN pairing. Per-device install steps: /docs/install-client.
Build & test (from source)
For development, or as an install fallback where no package is available:
cargo build --workspace # core, host, tray, shared client crates, Linux shell + session client, probe (Linux & macOS)
cargo test --workspace # unit + loopback + proptest + C ABI harness
cargo clippy --workspace --all-targets -- -D warnings
cargo fmt --all --check
cargo run -p loss-harness # FEC loss-resilience sweep (no network needed)
bash crates/punktfunk-core/tests/c/run.sh # standalone C-ABI link + round-trip proof
The C header regenerates from crates/punktfunk-core/src/abi.rs on every build (cbindgen via
build.rs) into include/punktfunk_core.h. The Apple, Android, and Windows clients have their own
toolchains (Xcode/swift build, Gradle, and cargo on the MSVC target) — see each client's README
and the docs site.
Layout
crates/
punktfunk-core/ protocol · FEC · pacing · crypto · QUIC control plane — the C ABI (lib + cdylib + staticlib)
punktfunk-host/ the host (Linux + Windows): virtual displays · capture · encode · input · GameStream · punktfunk/1 · mgmt
pf-client-core/ shared client plumbing (Linux + Windows): session pump · FFmpeg decode · audio · SDL3 gamepads · trust · discovery
pf-presenter/ Vulkan session presenter: SDL3 window · ash swapchain · frame present · input capture
pf-console-ui/ Skia console UI for the session client: gamepad shell · stats OSD · pairing · on-screen keyboard
pf-ffvk/ FFmpeg Vulkan hwcontext bindings (AVVkFrame) for Vulkan Video decode on the presenter's device
pf-driver-proto/ host ↔ pf-vdisplay driver contract: control IOCTLs + IDD-push frame transport (no_std)
punktfunk-tray/ host tray icon (Windows notification area / Linux StatusNotifierItem)
clients/
apple/ macOS / iOS / tvOS app (Swift · VideoToolbox · Metal · GameController)
linux/ Linux launcher shell (Rust · relm4 / GTK4 / libadwaita) — spawns the session client to stream
session/ punktfunk-session, the Vulkan streaming session (Rust · SDL3 · ash · Skia console UI) — also runs standalone (gamescope, Decky)
windows/ Windows desktop app (Rust · WinUI 3 · D3D11 · WASAPI · SDL3)
android/ Android phone + TV app (Kotlin · Rust JNI core · AMediaCodec · AAudio)
probe/ headless reference / measurement client for punktfunk/1
decky/ Steam Deck Decky plugin
web/ web console (TanStack) over the management API — status · devices · pairing · GPUs · performance · logs
api/openapi.json management-API OpenAPI spec (regenerated via `punktfunk-host openapi`, checked in)
packaging/ apt · rpm / COPR · Arch · Flatpak · Bazzite bootc image
docs-site/ public documentation site (Fumadocs) — https://docs.punktfunk.unom.io
include/punktfunk_core.h cbindgen-generated C header (checked in)
tools/ latency-probe · loss-harness (measurement)
ci/ CI container images (rust-ci · fedora-rpm)
Design invariants
- One core, linked everywhere. Protocol, FEC, and crypto live in
punktfunk-coreexactly once, exposed over a stable, versioned C ABI (punktfunk_abi_version(),PunktfunkConfigcarries its ownstruct_size). Every native client links the same core. - No async on the hot path. The per-frame pipeline uses native threads only;
tokio/quinnare gated behind the off-by-defaultquicfeature (control plane only). - Native client resolution, no scaling. Each session gets a virtual output at exactly the
client's WxH@Hz; each compositor keeps its own backend behind a shared
VirtualDisplaytrait. - FEC is the wall-breaker. GF(2⁸) (≤255 shards/block) for Moonlight compatibility; GF(2¹⁶)
(≤65535 shards/block, SIMD, O(n log n)) for
punktfunk/1to push past ~1 Gbps.
License
Licensed under either of
- Apache License, Version 2.0 (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)
at your option — SPDX-License-Identifier: MIT OR Apache-2.0.
Contribution
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions. See CONTRIBUTING.md.
Third-party components
punktfunk's own source is MIT/Apache-2.0. Shipped binaries additionally link third-party components
under their own (permissive) licenses — see THIRD-PARTY-NOTICES.txt
(regenerate with scripts/gen-third-party-notices.sh). The Windows host and client builds also
bundle FFmpeg under the LGPL v2.1+ (dynamically linked, replaceable DLLs; the license text and
notice ship in the installed licenses/ folder).
Trademarks
punktfunk is an independent project and is not affiliated with, endorsed by, or sponsored by NVIDIA, Microsoft, Sony, Valve, or the Moonlight project. "GameStream", "Moonlight", "Xbox", "DualSense", "DualShock", and "PlayStation" are trademarks of their respective owners and are used here only to describe interoperability.