[Bug tree-optimization/101936] New: 538.imagick_r -Ofast regression Kabylake

[Bug tree-optimization/101936] New: 538.imagick_r -Ofast regression Kabylake

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

            Bug ID: 101936
           Summary: 538.imagick_r -Ofast regression Kabylake
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
            Blocks: 26163
  Target Milestone: ---

When compiling with -Ofast -march=native, the benchmark is ~27% slower on
-march=skylake machine:
https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=32.507.0

I'm bisecting that right now on a Znver1 machine with -march=skylake set.


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/101936] [12 Regression] 538.imagick_r -Ofast regression Kabylake

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|538.imagick_r -Ofast        |[12 Regression]
                   |regression Kabylake         |538.imagick_r -Ofast
                   |                            |regression Kabylake
   Target Milestone|---                         |12.0
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=101935

[Bug tree-optimization/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |crazylht at gmail dot com
     Ever confirmed|0                           |1
            Summary|[12 Regression]             |[12 Regression]
                   |538.imagick_r -Ofast        |538.imagick_r -Ofast
                   |regression on Kabylake      |regression on Kabylake
                   |                            |since
                   |                            |g:872da9a6f664a06d73c987aa0
                   |                            |cb2e5b830158a10
   Last reconfirmed|                            |2021-08-16
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
Started with g:872da9a6f664a06d73c987aa0cb2e5b830158a10.

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

--- Comment #2 from Hongtao.liu <crazylht at gmail dot com> ---
I'll take a look at why there's the regression and revert my patch.

There's already another regression by this commit
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101929

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

--- Comment #3 from Hongtao.liu <crazylht at gmail dot com> ---
(In reply to Hongtao.liu from comment #2)
> I'll take a look at why there's the regression and revert my patch.
> 
I can't reproduce it on my skylake client, anyway i will revert my patch.
> There's already another regression by this commit
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101929

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by hongtao Liu <liuhongt@gcc.gnu.org>:

https://gcc.gnu.org/g:1db70e61a92978377a648bbd90e383859fc0126b

commit r12-3011-g1db70e61a92978377a648bbd90e383859fc0126b
Author: liuhongt <hongtao.liu@intel.com>
Date:   Tue Aug 17 17:29:06 2021 +0800

    Revert "Add the member integer_to_sse to processor_cost as a cost
simulation for movd/pinsrd. It will be used to calculate the cost of
vec_construct."

    This reverts commit 872da9a6f664a06d73c987aa0cb2e5b830158a10.

    PR target/101936
    PR target/101929

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

--- Comment #5 from Martin Liška <marxin at gcc dot gnu.org> ---
> https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=32.507.0
> 

Looking at LNT, the benchmark score is not back after the reversion of the
patch.
Maybe we face now PR101935?

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

--- Comment #6 from Hongtao.liu <crazylht at gmail dot com> ---
(In reply to Martin Liška from comment #5)
> > https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=32.507.0
> > 
> 
> Looking at LNT, the benchmark score is not back after the reversion of the
> patch.
> Maybe we face now PR101935?

probably.

[Bug target/101936] [12 Regression] 538.imagick_r -Ofast regression on Kabylake since g:872da9a6f664a06d73c987aa0cb2e5b830158a10

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101936

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

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

--- Comment #7 from Martin Liška <marxin at gcc dot gnu.org> ---
The performance is back, thus closing..