https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118959
Filip Kastl <pheeck at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |hjl at gcc dot gnu.org,
| |hubicka at gcc dot gnu.org
Summary|[15 Regression] 5-14% |[15 Regression] 5-14%
|slowdown of 400.perlbench |slowdown of 400.perlbench
| |since
| |r15-7400-gd3ff498c478ace
--- Comment #3 from Filip Kastl <pheeck at gcc dot gnu.org> ---
The same situation as in pr118957 (slowdown of povray).
Andrew is right. I have checked that the r15-7400-gd3ff498c478ace indeed
introduces the slowdown (I have measured a 12% slowdown).
I have looked at more perlbench results. I found that there are some
configurations where perlbench speeds up a bit (I have only seen them on Zen2
and they are mostly ~3%). Examples:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=788.10.0
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=288.10.0
The slowdowns are more numerous and bigger though.
Cc-ing H. J. Lu and Honza.
