[Bug target/112547] New: 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[Bug target/112547] New: 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

            Bug ID: 112547
           Summary: 9% exec time regression of 462.libquantum SPEC on AMD
                    zen4 CPU
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          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

As seen here

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=956.210.0

between commits

g:53010f6ff6dfbf7b (2023-11-05 02:42)

and

g:1a55724f7870719d (2023-11-06 02:27)

there is about 9% slowdown of execution time of the 2006SPEC 462.libquantum.

The test is run with -O2 and lto on an amd zen4 machine.

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization,
                   |                            |needs-bisection
                 CC|                            |haochen.jiang at intel dot com

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
The x86 changes from Haochen are the only relevant changes in that area I think
(but it's odd they affect -O2 generic tune).

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #2 from Haochen Jiang <haochen.jiang at intel dot com> ---
It is weird since I did not touch the tune.

Need a bisect to check that but I do not have a zen4 machine.

Could you try with this commit g:459866eaeec151e72aecd670695f014f4ec48588 to
see if the regression still occurs?

If that still occurs, a guess might be zmm vectorization is not enabled
corrected under some scenario.

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #3 from Haochen Jiang <haochen.jiang at intel dot com> ---
(In reply to Haochen Jiang from comment #2)
> It is weird since I did not touch the tune.
> 
> Need a bisect to check that but I do not have a zen4 machine.
> 
> Could you try with this commit g:459866eaeec151e72aecd670695f014f4ec48588 to
> see if the regression still occurs?
> 
> If that still occurs, a guess might be zmm vectorization is not enabled
> corrected under some scenario.

Sorry, should be if that regression disappeared, not occured.

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #4 from Haochen Jiang <haochen.jiang at intel dot com> ---
I checked the znver3 plot on the site, it seems that no regression occurs.

Since znver4 enabled AVX512, that is the reason why I guessed previously.

Could you also provide the option you ran with? I could not find where it hides
in the site. Thx!

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #5 from Filip Kastl <fkastl at suse dot cz> ---
Compile options are:
-O2 -g -flto=128

Configure options of GCC are:
--enable-languages=c,c++,fortran,rust,m2 --disable-bootstrap
--disable-libsanitizer --disable-multilib --enable-checking=release

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[liuhongt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

liuhongt at gcc dot gnu.org changed:

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

--- Comment #6 from liuhongt at gcc dot gnu.org ---
(In reply to Filip Kastl from comment #5)
> Compile options are:
> -O2 -g -flto=128
> 
> Configure options of GCC are:
> --enable-languages=c,c++,fortran,rust,m2 --disable-bootstrap
> --disable-libsanitizer --disable-multilib --enable-checking=release

hmm, so there's no avx512 in the cmdline options, haochen's commit shouldn't
have any impact in theory(the binary should be exact the same).
haochen, could you double check that?

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #7 from Haochen Jiang <haochen.jiang at intel dot com> ---
I have got a same binary w/ and w/o my commit with the options if nothing went
wrong.

Seems we need more investigation.

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

--- Comment #8 from Filip Kastl <fkastl at suse dot cz> ---
I've just ran the test on another zen4 machine. Between the originally
mentioned commits g:53010f6ff6dfbf7b and g:1a55724f7870719d there was only 1%
slowdown on this other machine. I guess this means that the 9% slowdown is
specific to the machine where we measure the data I sent.

Since there seems to be no reason why there should be a general zen4 slowdown
between the two commits and if there are no objections, I'll mark this bug as
RESOLVED WORKSFORME.

[Bug target/112547] 9% exec time regression of 462.libquantum SPEC on AMD zen4 CPU

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112547

Filip Kastl <fkastl at suse dot cz> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |WORKSFORME

--- Comment #9 from Filip Kastl <fkastl at suse dot cz> ---
Setting status to RESOLVED WORKSFORME