Revert "intel/fs: only avoid SIMD32 if strictly inferior in throughput"
authorMatt Turner <mattst88@gmail.com>
Thu, 10 Aug 2023 18:12:24 +0000 (14:12 -0400)
committerMarge Bot <emma+marge@anholt.net>
Wed, 16 Aug 2023 14:56:15 +0000 (14:56 +0000)
commitd142c845d05732f7858839493e0dba6a1fa5d6c0
tree6c255d22eea4711923d44911785cd73464559118
parentaebe58458611e0bb585a5bce8e16c1175783f3cc
Revert "intel/fs: only avoid SIMD32 if strictly inferior in throughput"

This reverts commit 6b494745be0900a67004d6f3e4b730c3cd67da79.

The logic is not entirely correct: the comparison is between two
static-analysis estimates of a dynamic system with variables that aren't
captured by the shader source, so using ">" will always have greater potential
to cause regressions whenever the performance difference between the two builds
is something not captured by the static model, no matter how much the model is
improved.

Reference: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9262
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24615>
src/intel/compiler/brw_fs.cpp