Skip to content

v3d: add support for load_workgroup_size

Karol Herbst requested to merge karolherbst/mesa:v3d/variable_workgroup into main

This wires up the compiler and the gallium driver bits.

Sadly, I don't see a way for actually exposing this via GL, as GL compute shaders require 1024 threads and ARB_compute_variable_group_size might be OpenGL only?

In any case, this will be required for OpenCL support as the workgroup size is variable by default (unless specified in the kernel).

Merge request reports