[Bug tree-optimization/43940] New: DOM doesn't propagate constants properly

[Bug tree-optimization/43940] New: DOM doesn't propagate constants properly

[rguenth at gcc dot gnu dot org]

For

int main()
{
  void *p = __builtin_malloc (4);
  if (p)
    {
      __builtin_free (p);
      p = 0;
    }
  __builtin_free (p);
  return 0;
}

DOM does not propagate p == 0 on the else path leaving

<bb 2>:
  p_2 = __builtin_malloc (4);
  if (p_2 != 0B)
    goto <bb 3>;
  else
    goto <bb 4>;

<bb 3>:
  __builtin_free (p_2);

<bb 4>:
  # p_1 = PHI <p_2(2), 0B(3)>
  __builtin_free (p_1);


VRP can do this.


-- 
           Summary: DOM doesn't propagate constants properly
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2010-07-24 21:44 -------
Confirmed.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-07-24 21:44:54
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[steven at gcc dot gnu dot org]

------- Comment #2 from steven at gcc dot gnu dot org  2010-07-24 22:04 -------
Investigating...


-- 

steven at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |steven at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2010-07-24 21:44:54         |2010-07-24 22:04:50
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[steven at gcc dot gnu dot org]

------- Comment #3 from steven at gcc dot gnu dot org  2010-07-25 10:29 -------
VRP does this with ASSERT_EXPRs.  DOM does not optimize this because bb4 has
two predecessors, and record_equivalences_from_incoming_edge only records from
a single predecessor.

This should probably be handled record_equivalences_from_phis by looking at the
value of the incoming argument, before calling operand_equal_for_phi_arg_p.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[steven at gcc dot gnu dot org]

------- Comment #4 from steven at gcc dot gnu dot org  2010-07-25 16:25 -------
cprop_into_successor_phis only propagates SSA_NAME_VALUEs, but not conditional
copies/constants. May be a better place to fix this bug.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[steven at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

--- Comment #5 from Steven Bosscher <steven at gcc dot gnu.org> 2012-03-18 22:48:49 UTC ---
With trunk today (r185508), on x86_64-linux at -O2 -fno-tree-vrp:

The .072t.ifcombine dump:
main ()
{
  void * p;

<bb 2>:
  p_2 = __builtin_malloc (4);
  if (p_2 != 0B)
    goto <bb 3>;
  else
    goto <bb 4>;

<bb 3>:
  __builtin_free (p_2);

<bb 4>:
  # p_1 = PHI <p_2(2), 0B(3)>
  __builtin_free (p_1);
  return 0;

}


And the .073t.phiopt1 dump:
main ()
{
  void * p;

<bb 2>:
  p_2 = __builtin_malloc (4);
  if (p_2 != 0B)
    goto <bb 3>;
  else
    goto <bb 4>;

<bb 3>:
  __builtin_free (p_2);

<bb 4>:
  # p_1 = PHI <0B(2), 0B(3)>
  __builtin_free (p_1);
  return 0;

}


And finally the 080t.dom1 dump:
main ()
{
  void * p;

<bb 2>:
  p_2 = __builtin_malloc (4);
  if (p_2 != 0B)
    goto <bb 3>; 
  else
    goto <bb 4>;

<bb 3>:
  __builtin_free (p_2);

<bb 4>:
  # p_1 = PHI <0B(2), 0B(3)>
  GIMPLE_NOP
  return 0;

}

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Depends on|                            |52424
         Resolution|                            |FIXED
   Target Milestone|---                         |4.7.0

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-03-19 09:34:58 UTC ---
I think this was fixed by

2012-02-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

        PR tree-optimization/52424
        * tree-ssa-dom.c (dom_opt_leave_block): Push a marker before
        calling dom_thread_across_edge.

[Bug tree-optimization/43940] DOM doesn't propagate constants properly

[stevenb.gcc at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43940

--- Comment #7 from stevenb.gcc at gmail dot com <stevenb.gcc at gmail dot com> 2012-03-19 09:51:41 UTC ---
> --- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-03-19 09:34:58 UTC ---
> I think this was fixed by
>
> 2012-02-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
>
>        PR tree-optimization/52424
>        * tree-ssa-dom.c (dom_opt_leave_block): Push a marker before
>        calling dom_thread_across_edge.

I think it was fixed by ifcombine. But in any case, yes, resolved AFAIC.