https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123126
Bug ID: 123126
Summary: [16 Regression] 6-10% slowdown of xalancbmk_r on
Zen{4,5} since r16-6104-gb5c64db0a49d46 with PGO
Product: gcc
Version: 16.0
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: pinskia at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-suse-linux
Target: x86_64-suse-linux
As seen here
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1286.367.0
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=958.367.0
there was a 6-10% exec time slowdown of the wrf xalancbmk_r SPEC 2017
benchmark when compiled with -Ofast -march=native -flto with PGO on a Zen4
machine and -O2 -march=x86-64-v3 -flto with PGO on a Zen5
machine. I bisected it to r16-6104-gb5c64db0a49d46.
b5c64db0a49d4653418457a520648c4305e01e75 is the first bad commit
commit b5c64db0a49d4653418457a520648c4305e01e75
Author: Andrew Pinski <[email protected]>
Date: Tue Nov 18 23:02:36 2025 -0800
ch: Improve copy header when the bbs are predicated as non-executed
[PR122734]
This is a regression against GCC 15. See the comparisons
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1267.367.0&plot.1=1333.367.0&plot.2=1296.367.0&plot.3=1353.367.0&plot.4=1286.367.0&;
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=952.367.0&plot.1=1006.367.0&plot.2=977.367.0&plot.3=1016.367.0&plot.4=1179.367.0&plot.5=958.367.0&;
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=26163
[Bug 26163] [meta-bug] missed optimization in SPEC (2k17, 2k and 2k6 and 95)
