Skip to content

intel/nir: Don't try to emit vector load_scratch instructions

Faith Ekstrand requested to merge gfxstrand/mesa:review/fs-no-vector-scratch into master

While we'er here, also add an assertion so that it's easier to catch accidentally emitting wide scratch loads and stores before we get into the back-end where we have no idea why it happened.

Fixes: 53bfcdee "intel/fs: Implement the new load/store_scratch..."

Merge request reports