https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118959
Bug ID: 118959
Summary: [15 Regression] 5-14% slowdown of 400.perlbench
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pheeck at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-linux
Target: x86_64-linux
As seen here
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=465.10.0
there was a 15% exec time slowdown of the 400.perlbench SPEC 2006 benchmark
between commits
r15-7319-ge2d32c81a993a2
r15-7400-gd3ff498c478ace
when run with -O2 -march=native -flto on an AMD Zen3 machine.
Similar slowdowns around that time:
8% on an AMD Zen4 machine with -O2 -march=native -flto
(although this graph is somewhat noisier)
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=993.10.0
??% on an AMD Zen4 machine with -Ofast -march=native -flto
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=992.10.0
5% on an AMD Zen4 machine with -Ofast -march=x86-64-v3 -flto
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1119.10.0
---
The fact that the Zen4 -Ofast graphs show a gradual slowdown over multiple
commits is interesting. This then maybe isn't a single regression? Not sure
how much we can do about slowdowns like this.
---
Here is a graph (for the 15% case) showing that this is a regression over GCC14
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.8=1039.10.0&plot.9=465.10.0&;
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)
