Skip to content

lima/gpir: A few optimizations for branches

Connor Abbott requested to merge cwabbott0/mesa:review/lima-gpir-branch-opt into main

This series for gpir fixes a few obvious inefficiencies with the code generated for multi-basic-block shaders, and also fixes regressions in code size for complicated single-block shaders due to the new register allocator. It also lays the foundation for how spilling will work.

Merge request reports