[Bug tree-optimization/123084] New: 4-10% slowdown of 544.nab_r on Aarch64 and Intel Ice Lake

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123084

            Bug ID: 123084
           Summary: 4-10% slowdown of 544.nab_r on Aarch64 and Intel Ice
                    Lake
           Product: gcc
           Version: 16.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, needs-bisection
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jmelcr at gcc dot gnu.org
                CC: pheeck at gcc dot gnu.org
            Blocks: 26163
  Target Milestone: ---

Hi,
the nab_r benchmark has slowed down by 4% to 10% with various flags:

Aarch64 + -O2 -flto:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=581.517.0

Aarch64 + -O2 -march=native -mtune=native -flto:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=582.517.0

Aarch64 + -Ofast:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=585.517.0

Intel Ice Lake + -Ofast -march=native -mtune=native:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=801.517.0

I have no way to bisect this, but the slowdowns intersect on commit range
09bece00d0e..cf22e30f0cf.

It doesn't appear to be a regression against GCC 15, as seen here:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=631.517.0&plot.1=679.517.0&plot.2=661.517.0&plot.3=651.517.0&plot.4=614.517.0&plot.5=598.517.0&plot.6=776.517.0&plot.7=875.517.0&plot.8=1045.517.0&plot.9=1235.517.0&plot.10=581.517.0&;

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=632.517.0&plot.1=681.517.0&plot.2=662.517.0&plot.3=654.517.0&plot.4=620.517.0&plot.5=607.517.0&plot.6=778.517.0&plot.7=880.517.0&plot.8=1076.517.0&plot.9=1230.517.0&plot.10=582.517.0&;

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=637.517.0&plot.1=682.517.0&plot.2=668.517.0&plot.3=645.517.0&plot.4=616.517.0&plot.5=602.517.0&plot.6=780.517.0&plot.7=882.517.0&plot.8=1074.517.0&plot.9=1233.517.0&plot.10=585.517.0&;

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=816.517.0&plot.1=848.517.0&plot.2=832.517.0&plot.3=911.517.0&plot.4=1093.517.0&plot.5=1200.517.0&plot.6=801.517.0&;


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)