[Bug tree-optimization/46297] New: [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[Bug tree-optimization/46297] New: [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

           Summary: [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with
                    -Os -fno-asynchronous-unwind-tables
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz


Output:
$ gfortran -Os -fno-asynchronous-unwind-tables 980701-0.f
$ ./a.out 
Aborted

valgrind output:
$ gfortran -Os -fno-asynchronous-unwind-tables 980701-0.f -g
$ valgrind -q ./a.out 
==3171== Conditional jump or move depends on uninitialised value(s)
==3171==    at 0x40078C: x_ (980701-0.f:67)
==3171== 
==3171== Conditional jump or move depends on uninitialised value(s)
==3171==    at 0x40078E: x_ (980701-0.f:67)
==3171== 
Aborted

The warning is for these instructions:
  400781:       f3 0f 10 00             movss  (%rax),%xmm0
  400785:       0f 2e 05 7c 02 00 00    ucomiss 0x27c(%rip),%xmm0        #
400a08 <options.0.1574+0x28>
  40078c:       7a 41                   jp     4007cf <x_+0x14b>
  40078e:       75 3f                   jne    4007cf <x_+0x14b>

Tested revisions:
r166272 - fail
r163636 - fail
r161659 - OK
4.5 r165781 - OK

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #1 from Zdenek Sojka <zsojka at seznam dot cz> 2010-11-04 02:16:26 UTC ---
Created attachment 22271
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22271
auto-reduced testcase

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.11.04 04:04:37
                 CC|                            |bernds at codesourcery dot
                   |                            |com, hjl.tools at gmail dot
                   |                            |com
   Target Milestone|---                         |4.6.0
     Ever Confirmed|0                           |1

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> 2010-11-04 04:04:37 UTC ---
It is caused by revision 162270:

http://gcc.gnu.org/ml/gcc-cvs/2010-07/msg00624.html

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at redhat dot com

--- Comment #3 from Jeffrey A. Law <law at redhat dot com> 2010-11-04 14:47:11 UTC ---
I think this might be a problem with handling auto-increment addressing modes
in the new optimization.  I'll be looking at it further later today.

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #4 from Jeffrey A. Law <law at redhat dot com> 2010-11-04 17:03:26 UTC ---
Testing a fix...

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

Bernd Schmidt <bernds at gcc dot gnu.org> changed:

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

--- Comment #5 from Bernd Schmidt <bernds at gcc dot gnu.org> 2010-11-15 12:30:26 UTC ---
Jeff, what have you found?

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #6 from Jeffrey A. Law <law at redhat dot com> 2010-11-15 14:13:55 UTC ---
On 11/15/10 05:40, bernds at gcc dot gnu.org wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297
>
> Bernd Schmidt<bernds at gcc dot gnu.org>  changed:
>
>             What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                   CC|                            |bernds at gcc dot gnu.org
>
> --- Comment #5 from Bernd Schmidt<bernds at gcc dot gnu.org>  2010-11-15 12:30:26 UTC ---
> Jeff, what have you found?
>
http://gcc.gnu.org/ml/gcc-patches/2010-11/msg00634.html

Waiting for review...

jeff

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #7 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-15 15:10:12 UTC ---
(In reply to comment #6)
> On 11/15/10 05:40, bernds at gcc dot gnu.org wrote:
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297
> >
> > Bernd Schmidt<bernds at gcc dot gnu.org>  changed:
> >
> >             What    |Removed                     |Added
> > ----------------------------------------------------------------------------
> >                   CC|                            |bernds at gcc dot gnu.org
> >
> > --- Comment #5 from Bernd Schmidt<bernds at gcc dot gnu.org>  2010-11-15 12:30:26 UTC ---
> > Jeff, what have you found?
> >
> http://gcc.gnu.org/ml/gcc-patches/2010-11/msg00634.html
> 
> Waiting for review...

You don't need explicit approval for this middle-end patch.

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #8 from Jeffrey A. Law <law at redhat dot com> 2010-11-15 18:05:11 UTC ---
On 11/15/10 08:14, rguenth at gcc dot gnu.org wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297
>
> --- Comment #7 from Richard Guenther<rguenth at gcc dot gnu.org>  2010-11-15 15:10:12 UTC ---
> (In reply to comment #6)
>> On 11/15/10 05:40, bernds at gcc dot gnu.org wrote:
>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297
>>>
>>> Bernd Schmidt<bernds at gcc dot gnu.org>   changed:
>>>
>>>              What    |Removed                     |Added
>>> ----------------------------------------------------------------------------
>>>                    CC|                            |bernds at gcc dot gnu.org
>>>
>>> --- Comment #5 from Bernd Schmidt<bernds at gcc dot gnu.org>   2010-11-15 12:30:26 UTC ---
>>> Jeff, what have you found?
>>>
>> http://gcc.gnu.org/ml/gcc-patches/2010-11/msg00634.html
>>
>> Waiting for review...
> You don't need explicit approval for this middle-end patch.
Noted.  More than happy to consider postreload.c part of the middle end :-)

I wouldn't mind hearing from Bernd to ensure that I'm marking the reg as 
being used in an unknown fashion correctly, but if I haven't heard from 
him in a few days, I'll go ahead and check it in.

jeff

[Bug tree-optimization/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #9 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-15 22:44:32 UTC ---
(In reply to comment #8)
> >>
> >> Waiting for review...
> > You don't need explicit approval for this middle-end patch.
> Noted.  More than happy to consider postreload.c part of the middle end :-)

Folklore says that everything that is not frontend is middle-end ;)

At least I'm using it that way when feeling comfortable to do so
(which usually excludes all RTL parts ...).

Richard.

[Bug middle-end/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
          Component|tree-optimization           |middle-end
         AssignedTo|unassigned at gcc dot       |law at redhat dot com
                   |gnu.org                     |

[Bug middle-end/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[law at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

--- Comment #10 from Jeffrey A. Law <law at gcc dot gnu.org> 2010-11-18 14:34:33 UTC ---
Author: law
Date: Thu Nov 18 14:34:29 2010
New Revision: 166905

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166905
Log:

    PR middle-end-optimization/46297
    * postreload.c (reload_combine_note_store): Deal with embedded
    side effects in MEM expressions.

    * gfortran.dg/pr46297.f: New test.



Added:
    trunk/gcc/testsuite/gfortran.dg/pr46297.f
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/postreload.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/46297] [4.6 Regression] gfortran.dg/g77/980701-0.f FAILs with -Os -fno-asynchronous-unwind-tables

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46297

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
                 CC|                            |jakub at gcc dot gnu.org
         Resolution|                            |FIXED

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-19 10:01:46 UTC ---
Fixed.