https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120751
Bug ID: 120751
Summary: [16 Regression] 10-15% slowdown of 454.calculix on
Zen4 and Zen5 since r16-1001-g0291f53f8d2343
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: pheeck at gcc dot gnu.org
CC: rguenth at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-pc-linux-gnu
Target: x86_64-pc-linux-gnu
As seen here
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1101.170.0
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1240.170.0
there was a 10-15% exec time slowdown of 454.calculix SPEC 2006
benchmark when run with -O2 -march=x86-64-v3 (or -march=native) -flto on a
Zen4/Zen5 machine.
I bisected it to r16-1001-g0291f53f8d2343.
0291f53f8d2343ca0d39589ebffc31d9c328d6ab is the first bad commit
commit 0291f53f8d2343ca0d39589ebffc31d9c328d6ab
Author: Richard Biener <rguent...@suse.de>
Date: Fri May 30 08:54:10 2025 +0200
tree-optimization/120457 - avoid lowering of some single-element interleave
The following makes sure we are not lowering single-element interleaving
schemes in a way that defeats load vectorizing later but allows the
VMAT_ELEMENTWISE fallback to be used.
PR tree-optimization/120457
* tree-vect-slp.cc (vect_lower_load_permutations): Implement
the same heuristics as load vectorization for single-element
interleaving that spans multiple vectors.
gcc/tree-vect-slp.cc | 9 +++++++++
1 file changed, 9 insertions(+)
This is a regression against GCC 15. See the comparison
here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1104.170.0&plot.1=1144.170.0&plot.2=1101.170.0&;
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)
