Mesa 22.0 Zink Speeds Up OpenGL-Over-Vulkan On CPUs

While there is already LLVMpipe Gallium3D for software acceleration of OpenGL on CPUs within Mesa, if wanting to increase the layers of abstraction you could also use Zink for OpenGL over Vulkan and by way of Lavapipe have that software accelerated on the CPU. With Mesa 22.0-devel, that route of Zink on CPUs is now faster.

Mike Blumenkrantz known for his prolific contributions to Zink while working under contract for Valve landed the latest optimizations. The improvement is on shader variants for compute uniform inlining and to enable this inlining by default when running on CPU-based drivers.

As Mike noted in the merge request that was pulled into Mesa 22.0-devel yesterday, it is "massively improving performance in a number of cases."

It will be interesting to see how well Zink works on CPU-based drivers and if it happens to uncover any areas of optimizations for itself or say LLVMpipe in comparison. I'll have up some fresh Zink GPU-based benchmarks shortly given all the gains made in recent months to this OpenGL-over-Vulkan implementation.