Skip to content

ac/nir/ngg: Support scalarized and packed 16-bit IO when lowering mesh shaders.

Timur Kristóf requested to merge Venemo/mesa:ac_nir_mesh_scalar_and_packed into main

This MR makes it possible for ac_nir_lower_ngg_ms to consume mesh shaders optimized by nir_opt_varyings, that have been lowered to scalar I/O and potentially pack two 16-bit outputs into a single 32-bit slot. There is also some code cleanup.

Merge request reports