[Bug tree-optimization/107499] New: 433.milc regressed by 6-8% on zen3 at -O2 -flto

[Bug tree-optimization/107499] New: 433.milc regressed by 6-8% on zen3 at -O2 -flto

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107499

            Bug ID: 107499
           Summary: 433.milc regressed by 6-8% on zen3 at -O2 -flto
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jamborm at gcc dot gnu.org
                CC: rguenth at gcc dot gnu.org
            Blocks: 26163
  Target Milestone: ---
              Host: x86_64-linux
            Target: x86_64-linux

LNT reports a 8.5% regression of 433.milc when compiled with -O2 -flto
on a zen3 machine:

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=469.70.0

On a similar machine I reproduced it as a 6.5% regression and bisected
it to 19295e8607d (tree-optimization/100756 - niter analysis and
folding).

Possibly related, possibly not:

- An intel machine results show something which might be a 2%
  regression around the same time at -O2 -flto -march=native.
  See https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=790.70.0 

- In the same time-frame there is also a 5% performance drop on zen2
  with -Ofast -flto -march=native, but those results have been rather
  flaky in the past (see PR 101296):
  https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=289.70.0

- there is a 11% regression reported on the same zen3 machine when
  using O2 flto -march=native but either I cannot reproduce it or it
  does not seem to be caused by the same commit.  I'll wait for more
  LNT results before investigating.  See
  https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=465.70.0


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)

[Bug tree-optimization/107499] [13 Regression] 433.milc regressed by 6-8% on zen3 at -O2 -flto

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107499

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|12.0                        |13.0
           Keywords|                            |missed-optimization
   Target Milestone|---                         |13.0
   Last reconfirmed|                            |2022-11-05
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
            Summary|433.milc regressed by 6-8%  |[13 Regression] 433.milc
                   |on zen3 at -O2 -flto        |regressed by 6-8% on zen3
                   |                            |at -O2 -flto

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
It should be possible to constrain the "fix" to niter analysis, it also affects
IVOPTs which might be the issue here.  I'll investigate during stage3.

[Bug tree-optimization/107499] [13 Regression] 433.milc regressed by 6-8% on zen3 at -O2 -flto

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107499

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
> it to 19295e8607d (tree-optimization/100756 - niter analysis and

g:19295e8607d

[Bug tree-optimization/107499] [13 Regression] 433.milc regressed by 6-8% on zen3 at -O2 -flto

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107499

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
>   I'll wait for more
>   LNT results before investigating.  See
>   https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=465.70.0

Apparently, the numbers are good again now.

[Bug tree-optimization/107499] [13 Regression] 433.milc regressed by 6-8% on zen3 at -O2 -flto

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107499

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Yep, the helene numbers remain but they started at a "high".  Assuming fixed.