https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126435
Bug ID: 126435
Summary: [17 Regression] 4-8% slowdown of blender_r on
Zen{2,3,4} since r17-2654-gfb0838bc1073aa
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: jmelcr at gcc dot gnu.org
CC: rguenth at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-suse-linux
Target: x86_64-suse-linux
Build: x86_64-suse-linux
As seen here
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=286.487.0
there was a 4% exec time slowdown of the blender_r SPEC 2017
benchmark when compiled with -Ofast -march=native -g -flto=128 PGO on an Zen2
machine. I bisected it to r17-2654-gfb0838bc1073aa.
fb0838bc1073aa5f22ec38e0cca2d70d625c6d71 is the first 'bad' commit
commit fb0838bc1073aa5f22ec38e0cca2d70d625c6d71
Author: Richard Biener <[email protected]>
Date: Wed Jul 15 10:37:57 2026 +0200
Support two-lane vector BB reductions without target support
This is a regression against GCC 16. See the comparison
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=326.487.0&plot.1=313.487.0&plot.2=350.487.0&plot.3=279.487.0&plot.4=402.487.0&plot.5=415.487.0&plot.6=775.487.0&plot.7=929.487.0&plot.8=1077.487.0&plot.9=1207.487.0&plot.10=1520.487.0&plot.11=286.487.0&;
There were also these blender_r slowdowns in the same timeframe (so probably
caused by the same commit):
6% Zen3 -Ofast -march=native -g -flto=128 PGO
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=477.487.0
6% Zen3 -Ofast -march=native -g -flto=128
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=475.487.0
8% Zen4 -Ofast -march=x86-64-v3 -g -flto=128
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1118.487.0
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=26163
[Bug 26163] [meta-bug] missed optimization in SPEC (2026, 2k17, 2k and 2k6 and
95)
