procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

rs6000: Set param_vect_partial_vector_usage as 1 for P10

Browse source 
Due to the unexpected performance on Power9 of those vector
with length instructions, we didn't enable vectorization with
partial vectors before.

Some recent testings show that they perform expectedly on
Power10 now.  This patch is to set
param_vect_partial_vector_usage as 1 for P10 or later.

The performance evaluation on the whole SPEC2017 with option
set power10/Ofast/unroll shows it can speed up 525.x264_r by
10.80% and 554.roms_r by 1.94%.

Bootstrapped/regtested on powerpc64le-linux-gnu P10.

gcc/ChangeLog:

	* config/rs6000/rs6000.c (rs6000_option_override_internal):
	Set param_vect_partial_vector_usage as 1 for Power10 and up
	by default.
This commit is contained in:
Kewen Lin 2020-11-26 03:20:40 -06:00
parent a386566118
commit f6702584aa
1 changed files with 7 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
gcc/config/rs6000/rs6000.c
View file

@ -4781,10 +4781,13 @@ rs6000_option_override_internal (bool global_init_p)
SET_OPTION_IF_UNSET (&global_options, &global_options_set,
param_max_completely_peeled_insns, 400);
/* Temporarily disable it for now since lxvl/stxvl on the default
supported hardware Power9 has unexpected performance behaviors. */
SET_OPTION_IF_UNSET (&global_options, &global_options_set,
param_vect_partial_vector_usage, 0);
/* The lxvl/stxvl instructions don't perform well before Power10. */
if (TARGET_POWER10)
SET_OPTION_IF_UNSET (&global_options, &global_options_set,
param_vect_partial_vector_usage, 1);
else
SET_OPTION_IF_UNSET (&global_options, &global_options_set,
param_vect_partial_vector_usage, 0);
/* Use the 'model' -fsched-pressure algorithm by default. */
SET_OPTION_IF_UNSET (&global_options, &global_options_set,