[Bug target/112922] New: 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[Bug target/112922] New: 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112922

            Bug ID: 112922
           Summary: 465.tonto from SPECFP 2006 fails train run on
                    Aarch64-linux with -O2 and -flto
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jamborm at gcc dot gnu.org
            Blocks: 26163
  Target Milestone: ---
              Host: aarch64-linux
            Target: aarch64-linux

At some point between r14-6211-g056cce412862f8 and
r14-6290-g9f0f7d802482a8 465.tonto started to fail its train run when
compiled with -O2 -g -flto and PGO on Aarch64-linux.

My Aarch64 machine is a bit slow so I'd like to ask someone with a
spare modern one to bisect this to an exact revision.  I do not
observe the failure on x86_64 (and so for now I have guessed its
component is "target").

The miscomparison is:

0056:  SCF Energy                  =      -407.833257
       SCF Energy                  =   -878703.963298
                                                    ^
0057:  Kinetic Energy              =       406.333744
       Kinetic Energy              =       426.858584
                                                    ^
0087:  Chi^2 in F                  =         4.143448
       Chi^2 in F                  =        19.253240
                                                    ^
0088:  Goodness of fit in F        =         2.035546
       Goodness of fit in F        =         4.387851
                                                    ^
0089:  R factor in F               =         0.046861
       R factor in F               =         0.096290
                                                    ^
0090:  Weighted R factor in F      =         0.054923
       Weighted R factor in F      =         0.118393
                                                    ^
0098:  Secondary extinction factor =        -0.000010
       Secondary extinction factor =        -0.000075
                                                    ^
0100:  Scale factor                =         1.010260
       Scale factor                =         1.023437
                                                    ^
0105:          0        0.140000        4.143448     -407.833257 *Damping on
               0        0.140000       19.253240  -878703.963298 *Damping on
                                               ^
0108:          1        0.140000        4.844793     -406.881314
               1        0.140000      105.264541 -1961055.460671


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 target/112922] [14 Regression] 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112922

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
            Summary|465.tonto from SPECFP 2006  |[14 Regression] 465.tonto
                   |fails train run on          |from SPECFP 2006 fails
                   |Aarch64-linux with -O2 and  |train run on Aarch64-linux
                   |-flto                       |with -O2 and -flto
                 CC|                            |pinskia at gcc dot gnu.org
   Target Milestone|---                         |14.0

[Bug target/112922] [14 Regression] 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112922

Alex Coplan <acoplan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |acoplan at gcc dot gnu.org,
                   |                            |rsandifo at gcc dot gnu.org

--- Comment #1 from Alex Coplan <acoplan at gcc dot gnu.org> ---
So I did some bisection on this, and indeed it seems to have started with
r14-6290-g9f0f7d802482a8958d6cdc72f1fe0c8549db2182 i.e.

commit 9f0f7d802482a8958d6cdc72f1fe0c8549db2182
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Thu Dec 7 19:41:19 2023

    aarch64: Add an early RA for strided registers

but then it seemed to get fixed shortly afterwards by
r14-6339-g8b5cd6c4519cc120badd2b35a9e30d4deb82c012 i.e.

commit 8b5cd6c4519cc120badd2b35a9e30d4deb82c012
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Fri Dec 8 16:27:40 2023

    aarch64: Some tweaks to the early-ra pass

CCing Richard S who can hopefully confirm if that change was expected to fix
correctness / wrong code issues.

[Bug target/112922] [14 Regression] 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112922

--- Comment #2 from Richard Sandiford <rsandifo at gcc dot gnu.org> ---
I don't remember there being a deliberate bug fix in that patch,
but there were some others later.  I suppose the optimistic case
is that this first went latent and then was fixed “properly”
afterwards.  But it could just be latent.

[Bug target/112922] [14 Regression] 465.tonto from SPECFP 2006 fails train run on Aarch64-linux with -O2 and -flto

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112922

Richard Sandiford <rsandifo at gcc dot gnu.org> changed:

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

--- Comment #3 from Richard Sandiford <rsandifo at gcc dot gnu.org> ---
Assume fixed by the patches for PR113295.  Please reopen if not.