← All resources

Inside Mesa 26.0's RADV RT improvements

Auto-discovered

Blog Advanced

pixelcluster

Visit resource ↗

Description

Inside Mesa 26.0's RADV RT improvements is a January 2026 technical blog post by pixelcluster on Mesa RADV Vulkan ray-tracing work for AMD GPUs.

It focuses on RT pipeline compilation and dispatch: adding proper shader function calls and calling conventions in ACO so any-hit and intersection shaders can compile separately instead of being inlined into one megashader. It covers Unreal Engine library linking pain, Ghostwire Tokyo RT speedups, parallel compile benefits, and a Lumen dispatch fix for 2D ray workgroups.

It is for advanced Mesa, Vulkan RT, and GPU compiler engineers who already understand ray-tracing pipelines and AMD shader compilation.

The post is a free long-form HTML article with diagrams and merge-request context on pixelcluster's GPU blog.

It is useful as a driver-side account of why separate any-hit compilation matters for compile stutter and traversal performance.