https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127647
Bug ID: 127647
Summary: slowdowns of SPEC 2006 tonto since
r17-4622-gbbf9808d787eac
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: rguenth at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-suse-linux
Target: x86_64-suse-linux
Build: x86_64-suse-linux
I am seeing many SPEC 2006 tonto slowdowns on various machines:
5% on znver2 with -Ofast -march=native:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=301.230.0
5% on znver2 with -Ofast:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=300.230.0
4% on znver3 with -Ofast:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=466.230.0
5% on znver3 with -Ofast -march=native:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=467.230.0
5% on znver3 with -Ofast -march=native PGO:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=468.230.0
5% on znver4 with -Ofast -march=native:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=965.230.0
5% on znver5 with -Ofast:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1243.230.0
7% on znver5 with -Ofast -march=native:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1246.230.0
7% on znver5 with -Ofast -march=native PGO:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1247.230.0
5% on Intel Ice Lake (3rd generation Xeon) with -Ofast:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=789.230.0
6% on Intel Ice Lake (3rd generation Xeon) with -Ofast -march=native PGO:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=791.230.0
I reproduced 3 of the slowdowns, on znver{3,4,5}, all of them bisect to
r17-4622-gbbf9808d787eac, so I suppose that's the case for all of them:
commit bbf9808d787eacf4464f2298d2f4265855b006bf
Author: Richard Biener <[email protected]>
Date: Wed Sep 23 11:04:39 2026 +0200
account for SSE to integer move cost for vec_to_scalar
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)
