nir_opt_algebraic: Don't shrink 64-bit bitwise ops if pack_split is going to be lowered
authorJesse Natalie <jenatali@microsoft.com>
Tue, 23 May 2023 20:22:42 +0000 (13:22 -0700)
committerMarge Bot <emma+marge@anholt.net>
Thu, 1 Jun 2023 00:36:10 +0000 (00:36 +0000)
commit6c62eaf22d72de4979fa6e9f90f71e6938e06560
treeab594467b0a1cb8b651969e8c09ad7fa702062ab
parent44b960a6453ec78d3cbb6624e6daaf8345d99dc4
nir_opt_algebraic: Don't shrink 64-bit bitwise ops if pack_split is going to be lowered

Otherwise this can cause optimizations to fight resulting in infinite
optimization loops with opt_algebraic, constant_folding, and copy_prop.

Fixes: 368be872 ("nir/algebraic: shrink 64-bit bitwise operations with 0/-1 constant half")
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23192>
src/compiler/nir/nir_opt_algebraic.py