https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125667
Bug ID: 125667
Summary: [17 Regression] 4-8% slowdown of cactusADM on Zen2
with PGO since r17-1273-g391ee229b737eb
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: jmelcr at gcc dot gnu.org
CC: pzheng 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=685.100.0
there was a 4-8% exec time slowdown of the cactusADM SPEC 2006
benchmark when compiled with -O2 -g -flto=128 with PGO on a Zen2
machine. I bisected it to r17-1273-g391ee229b737eb.
391ee229b737eb9a0cf3b705f41476f56d12d1e3 is the first bad commit
commit 391ee229b737eb9a0cf3b705f41476f56d12d1e3
Author: Pengxuan Zheng <[email protected]>
Date: Wed Apr 8 18:09:48 2026 -0700
cselim: Allow middle_bb to have more than one statement [PR124405]
This is about on par with GCC 15, a regression against older versions. See the
comparison
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=732.100.0&plot.1=734.100.0&plot.2=689.100.0&plot.3=746.100.0&plot.4=744.100.0&plot.5=762.100.0&plot.6=921.100.0&plot.7=1031.100.0&plot.8=1192.100.0&plot.9=1498.100.0&plot.10=685.100.0&;
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)
