https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126688
Bug ID: 126688
Summary: [17 Regression] 7% slowdown of fotonik3d_r on Zen5
since r17-2876-gf0bd50e8317926
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: roger at nextmovesoftware dot com
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=1254.527.0
there was a 7% exec time slowdown of the fotonik3d_r SPEC 2017
benchmark when compiled with -O2 -march=native -g on an Zen5
machine. I bisected it to r17-2876-gf0bd50e8317926.
commit f0bd50e831792692ffcedf288b312167a9339260 (HEAD)
Author: Roger Sayle <[email protected]>
Date: Sat Aug 1 17:51:02 2026 +0100
x86 SSE: Improved vector initialization/construction.
This is a regression against GCC 16. See the comparison
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1279.527.0&plot.1=1340.527.0&plot.2=1307.527.0&plot.3=1362.527.0&plot.4=1563.527.0&plot.5=1254.527.0&;
I tested the patch linked in pr126619, sadly it doesn't fix the issue.
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)
