Three Layers of the Linux Graphics Stack Get Major Updates in One Week

Firefox’s hardware video decoding has long relied on VA-API. While this interface worked reliably on Intel GPUs, NVIDIA proprietary driver support was always incomplete, and coverage for ARM-based embedded GPUs was nearly nonexistent. The result was a fragmented experience: the same 1080p stream could be GPU-accelerated on one machine and software-decoded on another, depending solely on vendor.

The fix that landed this week comes from NVIDIA engineer Tymur Boiko, with integration work by Red Hat’s Martin Stransky. By routing hardware decoding through the Vulkan Video API, Firefox gains a vendor-neutral path that works uniformly across NVIDIA, AMD, Intel, and Qualcomm — including ARM mobile and embedded devices. Vulkan is also protocol-agnostic: it operates on both X11 and Wayland without modification.

Firefox 153.0 is targeted for release on July 21, 2026, assuming no late-breaking issues.

The most technically significant change in Wayland Protocols 1.49 is the improved multi-GPU support for the linux-dmabuf-v1 protocol. KDE developer Xaver Hugl began this work in 2023, and it has finally landed. DMA-BUF is the kernel mechanism that transfers frame buffers between the GPU and the display subsystem without copying through CPU memory.

In multi-GPU systems — such as a laptop with both an integrated Intel GPU and a discrete NVIDIA or AMD card — the compositor previously had no reliable way to communicate to a client whether a DMA-BUF buffer import from a secondary device had succeeded. Clients rendering on a GPU other than the compositor’s main device had to guess, leading to silent failures and expensive fallback paths. The updated protocol lets compositors explicitly advertise every supported device on the system, together with the buffer formats and modifiers they can actually sample from, removing the guesswork entirely.

Version 1.49 also adds Windows BT.2100 image description request support to the color-management-v1 protocol, allowing Wayland compositors to apply correct handling for HDR content that carries Windows-compatible BT.2100 metadata. A third addition by Hugl is the experimental xx-fractional-scale-v2 protocol, which provides a cleaner coordinate space for fractional display scaling — an increasingly important capability as high-DPI panels become standard.

The VK_GOOGLE_display_timing extension merged into Mesa 26.2-devel this week specifically targets VK_KHR_display — Vulkan’s direct display mode, where an application bypasses the desktop compositor and talks to the display hardware directly. This path is used by VR runtimes, kiosk applications, and game exclusive fullscreen modes.

Developer Mario Kleiner’s implementation, which builds on prior Mesa work dating back to 2018, lets applications query the precise timestamp of the previous frame’s actual appearance on screen. Armed with that ground-truth timestamp, a rendering loop can eliminate minor frame-pacing drift without relying on predictions. The practical benefit is invisible during ordinary desktop use, but meaningful in two scenarios: VR, where frame timing errors contribute directly to motion sickness, and competitive gaming, where consistent frame delivery affects input-to-display latency in ways that experienced players notice.

The extension immediately benefits Intel ANV, AMD RADV, PowerVR, Qualcomm Turnip, and V3DV drivers. It complements the more powerful VK_EXT_present_timing that landed in Mesa 26.1, enabling existing applications already using the Google extension to benefit without requiring a porting effort.