https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126372
Bug ID: 126372
Summary: [17 Regression] 6% slowdown of calculix on Zen2 since
r17-2478-g3adb3325954177
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=300.170.0
there was a 6% exec time slowdown of the calculix SPEC 2006
benchmark when compiled with -Ofast -g on an Zen2
machine. I bisected it to r17-2478-g3adb3325954177.
3adb3325954177939d59031055fc08a81ae21be3 is the first 'bad' commit
commit 3adb3325954177939d59031055fc08a81ae21be3
Author: Richard Biener <[email protected]>
Date: Wed Jul 15 11:33:50 2026 +0200
Improve BB vectorization of reductions
This is a regression against GCC 16. See the comparison
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=329.170.0&plot.1=316.170.0&plot.2=337.170.0&plot.3=267.170.0&plot.4=405.170.0&plot.5=414.170.0&plot.6=754.170.0&plot.7=922.170.0&plot.8=1062.170.0&plot.9=1215.170.0&plot.10=1512.170.0&plot.11=300.170.0&;
There was also this calculix slowdown on aarch64 Neoverse N1, likely caused by
the same commit:
-Ofast -march=native -g PGO
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=578.170.0
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)
