Skip to content

gallivm: use fdiv rather than fmul * scale for unsigned_norm_to_float

Mike Blumenkrantz requested to merge zmike/mesa:llvm-maffs into main

What does this MR do and why?

gallivm: use fdiv rather than fmul * scale for unsigned_norm_to_float

this avoids pre-rounding the divisor and gives slightly more accurate
results, mainly in vertex fetches

cc: mesa-stable

Merge request reports