[Bug rtl-optimization/57425] New: RTL alias analysis unprepared to handle stack slot sharing

[Bug rtl-optimization/57425] New: RTL alias analysis unprepared to handle stack slot sharing

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

            Bug ID: 57425
           Summary: RTL alias analysis unprepared to handle stack slot
                    sharing
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: major
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: amylaar at gcc dot gnu.org

Created attachment 30198
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30198&action=edit
testcase variant using a int / long union

I see a failures for gcc.dg/torture/pr25654.c for avr for -Os, -O2 and higher
optimization levels.  The .172r.cse1 debugging dump still looks OK, but the
.173r.fwprop1 dump shows that the second memory read to compute sum
has been eliminated by reusing the value of the first one, despite the store
that happened in the mean time.

AFAICT, this patch:
http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=188667
has removed the infrastructure that was put in place to fix PR 25654, without
putting any in any replacement.

Now, what makes the avr different from the majority of tier1 targets is that
short and int are the same size.

If I change the testcase a bit to use int/long instead of short/int, it starts
failing on i686-pc-linux-gnu.  This is a 4.8 regression relative to 4.7 .
A variant of the testcase without unions also fails.  This is again at -Os /
-O2
and higher optimization levels.

[Bug rtl-optimization/57425] RTL alias analysis unprepared to handle stack slot sharing

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #1 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
Created attachment 30199
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30199&action=edit
testcase variant using int / long arrays

[Bug rtl-optimization/57425] RTL alias analysis unprepared to handle stack slot sharing

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #2 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
Created attachment 30200
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30200&action=edit
Test case variant using long / long long arrays

This, too, fails on i686-pc-linux-gnu.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-05-27
                 CC|                            |ebotcazou at gcc dot gnu.org
   Target Milestone|---                         |4.8.1
            Summary|RTL alias analysis          |[4.8/4.9 regression] RTL
                   |unprepared to handle stack  |alias analysis unprepared
                   |slot sharing                |to handle stack slot
                   |                            |sharing
     Ever confirmed|0                           |1

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Confirmed, but the .172r.cse1 dump already looks suspicious to me.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note that I only recently fixed a wrong alias query from using true_dependence
to anti_dependence.  Similar issues may exist elsewhere.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Note that I only recently fixed a wrong alias query from using
> true_dependence to anti_dependence.  Similar issues may exist elsewhere.

That would point to cselib_invalidate_mem then.  It uses canon_true_dependence
but here we have an anti-dependence (WAR).

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.1                       |4.8.2

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.1 has been released.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #30310|0                           |1
        is obsolete|                            |

--- Comment #8 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
Created attachment 30311
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30311&action=edit
patch with cselib / testsuite bits added

I'm currently testing this patch.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.2                       |4.8.3

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.2 has been released.

[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

Mikael Pettersson <mikpelinux at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpelinux at gmail dot com

--- Comment #10 from Mikael Pettersson <mikpelinux at gmail dot com> ---
I believe this is fixed on trunk since June, by r200133 + r200177 + r200241.

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.9.0
            Summary|[4.8/4.9 regression] RTL    |[4.8 Regression] RTL alias
                   |alias analysis unprepared   |analysis unprepared to
                   |to handle stack slot        |handle stack slot sharing
                   |sharing                     |
      Known to fail|                            |4.8.2
           Severity|major                       |normal

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed on trunk.

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

Bill Schmidt <wschmidt at gcc dot gnu.org> changed:

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

--- Comment #12 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
(In reply to Eric Botcazou from comment #5)
> > Note that I only recently fixed a wrong alias query from using
> > true_dependence to anti_dependence.  Similar issues may exist elsewhere.
> 
> That would point to cselib_invalidate_mem then.  It uses
> canon_true_dependence but here we have an anti-dependence (WAR).

I believe I've run into this problem in 4.8 as well when compiling some code
for powerpc64le-linux-gnu (so for now that is on ibm/gcc-4_8-branch).  At least
I'm seeing invalid commoning involving memory references in cselib, and the
problem doesn't reproduce on trunk.

Does anyone have a patch in progress for backporting this fix to 4.8?

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #13 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Created attachment 32343
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32343&action=edit
4.8 backport of Joern's fixes

This is the backport of Joern's fixes to 4.8 I did last summer.  Tested w/o
regressions on x86_64, sparc64, powerpc64, armv5tel, and m68k.

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #14 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Hi Mikael,

Indeed, your patch fixes my problem as well.  Do you have plans to push this
upstream?  It would be helpful if we could merge this from gcc-4_8-branch into
ibm/gcc-4_8-branch for an upcoming delivery.

Thanks!
Bill

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #15 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Bill,

I can submit the patch on Saturday (too busy before then), but I don't have
commit rights so someone else will have to commit it (assuming it's approved).

/Mikael

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #16 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Mikael, thanks very much.  I'll be happy to commit on your behalf once the
patch is approved.

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #17 from Mikael Pettersson <mikpelinux at gmail dot com> ---
The backport patch has now been submitted:
http://gcc.gnu.org/ml/gcc-patches/2014-03/msg00758.html

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #18 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Bill, the backport patch has now been approved:
http://gcc.gnu.org/ml/gcc-patches/2014-03/msg00792.html

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #19 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Author: wschmidt
Date: Mon Mar 17 15:31:43 2014
New Revision: 208620

URL: http://gcc.gnu.org/viewcvs?rev=208620&root=gcc&view=rev
Log:
gcc/

2014-03-17  Mikael Pettersson  <mikpelinux@gmail.com>
        Committed by Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

    Backport from mainline:

    2013-06-20  Joern Rennecke <joern.rennecke@embecosm.com>

    PR rtl-optimization/57425
    PR rtl-optimization/57569
    * alias.c (write_dependence_p): Remove parameters mem_mode and
    canon_mem_addr.  Add parameters x_mode, x_addr and x_canonicalized.
    Changed all callers.
    (canon_anti_dependence): Get comments and semantics in sync.
    Add parameter mem_canonicalized.  Changed all callers.
    * rtl.h (canon_anti_dependence): Update prototype.

    2013-06-16  Joern Rennecke <joern.rennecke@embecosm.com>

    PR rtl-optimization/57425
    PR rtl-optimization/57569
    * alias.c (write_dependence_p): Add new parameters mem_mode,
    canon_mem_addr and mem_canonicalized.  Change type of writep to bool.
    Changed all callers.
    (canon_anti_dependence): New function.
    * cse.c (check_dependence): Use canon_anti_dependence.
    * cselib.c (cselib_invalidate_mem): Likewise.
    * rtl.h (canon_anti_dependence): Declare.

gcc/testsuite/

2014-03-17  Mikael Pettersson  <mikpelinux@gmail.com>
        Committed by Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

    Backport from mainline:

    2013-06-16  Joern Rennecke <joern.rennecke@embecosm.com>

    PR rtl-optimization/57425
    PR rtl-optimization/57569
    * gcc.dg/torture/pr57425-1.c, gcc.dg/torture/pr57425-2.c: New files.
    * gcc.dg/torture/pr57425-3.c, gcc.dg/torture/pr57569.c: Likewise.


Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/torture/pr57425-1.c
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/torture/pr57425-2.c
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/torture/pr57425-3.c
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/torture/pr57569.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/alias.c
    branches/gcc-4_8-branch/gcc/cse.c
    branches/gcc-4_8-branch/gcc/cselib.c
    branches/gcc-4_8-branch/gcc/rtl.h
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/57425] [4.8 Regression] RTL alias analysis unprepared to handle stack slot sharing

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

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

--- Comment #20 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Fixed everywhere now.