RFC: Native Vulkan 1.3 & HDR10/scRGB Renderer Backend (smithay-renderer-vulkan)

[123llyrr]

Hi everyone,

Over the past months of building Clarion (a modern Wayland compositor), one of our major focuses has been bringing a first-class, production-grade native Vulkan 1.3 rendering backendwith full HDR10 & scRGB color pipelines to the Smithay ecosystem.

Knowing that native Vulkan has been an open goal for Smithay (and high on the wishlist for desktop projects like COSMIC), we have fully decoupled this rendering pipeline into an independent, dual-licensed (MIT OR Apache-2.0) crate:

👉 Repository: https://github.com/123llyrr/smithay-renderer-vulkan

What's Implemented & Tested

1. Smithay Trait Compliance:

   • Implements Renderer, ImportMemWl, ImportDmaWl, ImportDma, ExportMem, and Texture.
   • Native integration with DRM and nested Winit backends without forcing EGL fallback.

2. Modern Vulkan 1.3 Architecture:

   • Dynamic Rendering** (VK_KHR_dynamic_rendering / Vulkan 1.3 Core): Eliminates the overhead of legacy RenderPasses and Framebuffers.
   • Zero-Copy DMA-BUF Import: Direct DRM format modifier queries (VK_EXT_image_drm_format_modifier) matching linux_dmabuf_v1.
   • Explicit Sync & Timeline Semaphores: Native VK_KHR_timeline_semaphore + SYNC_FD export (wp_linux_drm_syncobj_v1 ready).

3. Complete HDR & Color Management Pipeline:

   • Full support for HDR10 (PQ / BT.2020) and scRGB (Linear FP16)framebuffers.
   • Dynamic transfer functions and color matrix conversions (sRGB, DCI-P3, Display-P3, BT.2020).
   • Tone-mapping kernels and automatic disjoint damage region scissoring.

4. Hardware Verified:

   • Verified on Intel Iris Xe / Arc, AMD RDNA, and NVIDIA drivers (running on real DRM render nodes).
   • Includes a minimal runnable example (cargo run --example vulkan_minimal).

---

Upstream Roadmap & Discussion

We would love to discuss with @Drakulix and the community how you would prefer to integrate this capability:

• Option A (External Crate in Smithay Ecosystem)**: Keep smithay-renderer-vulkan maintained as a dedicated ecosystem crate (like smithay-clipboard or cosmic-protocols), ready for immediate adoption by compositors (cosmic-comp, niri, etc.).
• Option B (Upstream into Smithay Core): Incrementally upstream modules into smithay::backend::renderer::vulkan.

Looking forward to your thoughts, feedback, and architecture review!

[Drakulix]

While I do not oppose community contributions for this topic, upstream work on a vulkan-renderer has already started and is expected to adjust the renderer-api to be more vulkan-friendly: #2165

Given this is a full "code-dump" for the feature and not a small demo + ask for initial feedback, I have to ask: Was this largely written with LLM-assistance?

[123llyrr]

Yes, it is an auxiliary development. I have done hardware testing.

…

------------------ Original ------------------ From: Victoria Brekenfeld ***@***.***> Date: Wed,Oct 21,2026 7:52 PM To: Smithay/smithay ***@***.***> Cc: 123llyrr ***@***.***>, Author ***@***.***> Subject: Re: [Smithay/smithay] RFC: Native Vulkan 1.3 & HDR10/scRGB Renderer Backend (smithay-renderer-vulkan) (Discussion #2234) While I do not oppose community contributions for this topic, upstream work on a vulkan-renderer has already started and is expected to adjust the renderer-api to be more vulkan-friendly: #2165 Given this is a full "code-dump" for the feature and not a small demo + ask for initial feedback, I have to ask: Was this largely written with LLM-assistance? — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS and Android. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[123llyrr]

Hi @maintainer, thank you for the honest and thoughtful response!        To answer your question directly: Yes, this implementation was developed with LLM assistance, directed and    tested by me on physical hardware (Intel Iris Xe + NVIDIA RTX 2050 mobile).        I completely understand your concern. The last thing I want to do is create a maintenance burden for upstream   with an unsolicited "code dump" — especially when #2165 is actively reworking the `Renderer` API to be more       Vulkan-friendly.        Because I know upstream review bandwidth is precious, my intention was never to force this into Smithay's       core tree:     1. I have packaged and published this independently on crates.io as [`smithay-renderer-   vulkan`](https://crates.io/crates/smithay-renderer-vulkan) so compositors can experiment with native Vulkan 1.3   today without burdening Smithay's core maintenance.     2. I genuinely love the Smithay ecosystem, and I wanted to share empirical evidence (DMA-BUF zero-copy,         timeline semaphores, sub-millisecond frame delivery) that Vulkan composition brings significant real-world        latency, thermal, and VRAM benefits to Wayland desktops.        Please feel free to close this PR if it doesn't align with the current architecture. If any part of the         standalone crate or the benchmarks can serve as a reference or test case for #2165, that would already be an      honor.        Thank you for all your incredible work on Smithay!