OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

intel/compiler: Postpone most int64 lowering to brw_postprocess_nir

Browse Source 
Float conversions continue to be lowered early at the same time as
nir_lower_doubles, which we run early so we don't have to run it for
every shader key variant.  However, all other int64 lowering is now
done late, after nir_opt_load_store_vectorize(), allowing it to
comprehend basic arithmetic on 64-bit addresses.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23064>
This commit is contained in:
Kenneth Graunke
2023-05-12 02:21:03 -07:00
committed by Marge Bot
parent d764c03d5d
commit 78a195f252
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/intel/compiler/brw_nir.c
View File

@@ -982,7 +982,7 @@ brw_preprocess_nir(const struct brw_compiler *compiler, nir_shader *nir,
brw_nir_optimize(nir, compiler, is_scalar);
OPT(nir_lower_doubles, opts->softfp64, nir->options->lower_doubles_options);
if (OPT(nir_lower_int64)) {
if (OPT(nir_lower_int64_float_conversions)) {
OPT(nir_opt_algebraic);
OPT(nir_lower_doubles, opts->softfp64,
nir->options->lower_doubles_options);