[Bug target/112916] New: [14 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[Bug target/112916] New: [14 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112916

            Bug ID: 112916
           Summary: [14 Regression] ~4-7% exec time regression of 433.milc
                    on AMD Zen2
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: needs-bisection
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: fkastl at suse dot cz
  Target Milestone: ---
              Host: x86_64-linux
            Target: x86_64-linux

Between commits

g:7758cb4b53e8a336

and

g:cff1fa6625d1273f

there's a slowdown of 433.milc 2006 SPEC. 7% regression seen on the graph here
(Zen2 machine):

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=288.70.0

The options are:
-Ofast -march=native -mtune=native LTO and PGO

I've also replicated this on another Zen2 machine and got 4% slowdown. I
haven't seen this slowdown on other architectures.

[Bug target/112916] [14 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112916

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
The graph is very noisy though and there's some gaps without data?

[Bug target/112916] [14 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112916

--- Comment #2 from Filip Kastl <fkastl at suse dot cz> ---
Sometimes there are gaps in the data our team gathers, unfortunately. One cause
can be that upstream GCC temporarily fails to build a benchmark. That shouldn't
matter in this case I think. The last gap for this benchmark was in September.

[Bug target/112916] [14 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112916

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
                 CC|                            |law at gcc dot gnu.org

[Bug target/112916] [14/15 Regression] ~4-7% exec time regression of 433.milc on AMD Zen2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112916

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|14.0                        |14.2

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 14.1 is being released, retargeting bugs to GCC 14.2.