https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118852
Bug ID: 118852
Summary: Train run of 502.gcc_r compiled with -Ofast
-fprofile-generate -march=x86_64-v3 fails
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: jamborm at gcc dot gnu.org
CC: tnfchris at gcc dot gnu.org
Blocks: 26163
Target Milestone: ---
Host: x86_64-linux-gnu
Target: x86_64-linux-gnu
When building SPEC 2017 benchmark 502.gcc_r with master revision
cfdb961588b and compiler options
-Ofast -g -fprofile-generate -fgnu89-inline -fno-strict-aliasing
-march=x86-64-v3
(or with -march=znver3, -march=znver4 or -march=znver5), the resulting
binary then "fails" (i.e. exits with a non-zero code during SPEC
training, IIUC).
This is without LTO and let me stress that I use -fno-strict-aliasing
so this is should not be a manifestation of PR 115256. Also,
-fprofile-generate seems to be important, just running the train run
with benchmark built without it works fine.
Unfortunately it is not easy to bisect this because, first, building
of the benchmark started failing because of dominator violations with
r15-6807-g68326d5d1a593d (Alex Coplan: vect: Force alignment peeling
to vectorize more early break loops [PR118211]) and when this is fixed
with r15-6809-gf1c6789ab6c544 (Tamar Christina: vect: Fix dominators
when adding a guard to skip the vector loop [PR118211]), this failure
ensues.
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)
