[Bug target/112879] New: [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[Bug target/112879] New: [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[fkastl at suse dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112879

            Bug ID: 112879
           Summary: [14 Regression] 4% exec time regression of 525.x264_r
                    on AMD Zen4
           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-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

There's a 4% slowdown of 525.x264_r on AMD Zen4 between commits

g:af7fa3135b6b046f

g:e85c596ae2d1e5f5

Here is a plot showing this regression:

https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=958.377.0

I've seen this regression on two different Zen4 machines but haven't observed
it on machines with different architecture.

[Bug target/112879] [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112879

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0
                 CC|                            |liuhongt at gcc dot gnu.org

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Possibly the reduc_* support in the backend now BB vectorizes unprofitable
cases.

[Bug target/112879] [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[liuhongt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112879

--- Comment #2 from Hongtao Liu <liuhongt at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> Possibly the reduc_* support in the backend now BB vectorizes unprofitable
> cases.

maybe zmm reduction, and we should have reasonable cost for reduction
vec_to_scalar.

[Bug target/112879] [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[pheeck at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112879

Filip Kastl <pheeck at gcc dot gnu.org> changed:

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

--- Comment #3 from Filip Kastl <pheeck at gcc dot gnu.org> ---
This slowdown is also reported in pr113600 along with two other configurations
where the slowdown occurs. It could be possible that these three configurations
are different slowdowns with different causes. I find that unlikely though.
I've even bisected this LTO PGO slowdown to the same commit as Martin Jambor
did with the PGO slowdown (r14-5603-g2b59e2b4dff421).

So I'm marking this as a duplicate of pr113600.

*** This bug has been marked as a duplicate of bug 113600 ***

[Bug target/112879] [14 Regression] 4% exec time regression of 525.x264_r on AMD Zen4

[pheeck at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112879

Filip Kastl <pheeck at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|needs-bisection             |

--- Comment #4 from Filip Kastl <pheeck at gcc dot gnu.org> ---
This slowdown is also reported in pr113600 along with two other configurations
where the slowdown occurs. It could be possible that these three configurations
are different slowdowns with different causes. I find that unlikely though.
I've even bisected this LTO PGO slowdown to the same commit as Martin Jambor
did with the PGO slowdown (r14-5603-g2b59e2b4dff421).

So I'm marking this as a duplicate of pr113600.