[Bug tree-optimization/127736] New: [17 Regression] 15% slowdown of SPEC 2006 milc on aarch64 since r17-4808-ga9b19ded78ca2d

[jmelcr at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127736

            Bug ID: 127736
           Summary: [17 Regression] 15% slowdown of SPEC 2006 milc on
                    aarch64 since r17-4808-ga9b19ded78ca2d
           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: aarch64-suse-linux
            Target: aarch64-suse-linux
             Build: aarch64-suse-linux

I am seeing a 15% slowdown of SPEC 2006 milc on aarch64 (Neoverse N1) when
compiling with -O2 -march=native:

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=580.70.0

The slowdown bisects to r17-4808-ga9b19ded78ca2d:
commit a9b19ded78ca2dd8f66b355b05ebbbb149c2d320
Author: Richard Biener <rguent...@suse.de>
Date:   Thu Oct 1 09:21:30 2026 +0200

    tree-optimization/127686 - mis-applied -O2 vector cost model heuristic

This is a regression against GCC 16:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1042.70.0&plot.1=1224.70.0&plot.2=1521.70.0&plot.3=580.70.0&;

The blamed commit is a fix for PR127686, sadly it did have fallout.


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)