projects / mesa.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: ae02622) | patch
nir/flrp: Lower flrp(a, b, #c) differently
authorIan Romanick <ian.d.romanick@intel.com>
Sat, 18 Aug 2018 23:53:55 +0000 (16:53 -0700)
committerIan Romanick <ian.d.romanick@intel.com>
Tue, 7 May 2019 05:52:29 +0000 (22:52 -0700)
commitc995d1ca3a3f14c2e6823ecdad90e7bb03e70c41
tree598281a5007dcb5192719a7e7b1d093f54fc397ftree
parentae02622d8fd34d62aeeb0a1b6c8c71bb16a5f32dcommit | diff
nir/flrp: Lower flrp(a, b, #c) differently

This doesn't help on Intel GPUs now because we always take the
"always_precise" path first.  It may help on other GPUs, and it does
prevent a bunch of regressions in "intel/compiler: Don't always require
precise lowering of flrp".

Reviewed-by: Matt Turner <mattst88@gmail.com>
src/compiler/nir/nir_lower_flrp.c diff | blob | history