[Bug tree-optimization/46217] New: [4.6 Regression] store to volatile is removed

[Bug tree-optimization/46217] New: [4.6 Regression] store to volatile is removed

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

           Summary: [4.6 Regression] store to volatile is removed
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: blocker
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: pinskia@gcc.gnu.org
                CC: rguenth@gcc.gnu.org


Take:
int foo(int a)
{
  int t;
  *(volatile int *)&t = a;
}
--- CUT ---

We remove the store to t now.

I think this was caused by MEM_REF.

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
   Target Milestone|---                         |4.6.0

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2010.10.29 12:58:31
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1
           Severity|blocker                     |normal

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-10-29 12:58:31 UTC ---
I will have a look.  But I think that automatic volatiles are odd anyway ;)

The problem is that DCE doesn't mark the statement as always needed.

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-10-29 13:03:07 UTC ---
Actually we rewrite it into SSA form where we have no way to convince the
operand scanner to preserve the volatileness.  I guess whoever uses automatic
volatiles deserves to get stack memory.

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-02 16:08:19 UTC ---
I have a patch.

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-03 11:05:43 UTC ---
Author: rguenth
Date: Wed Nov  3 11:05:39 2010
New Revision: 166239

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166239
Log:
2010-11-03  Richard Guenther  <rguenther@suse.de>

    PR tree-optimization/46217
    * tree-ssa.c (non_rewritable_mem_ref_base): Check if volatile
    flag matches.
    (execute_update_addresses_taken): Likewise.

    * gcc.dg/pr46217.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/pr46217.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa.c

[Bug tree-optimization/46217] [4.6 Regression] store to volatile is removed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46217

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

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

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-03 11:06:26 UTC ---
Fixed.