[Bug tree-optimization/19788] New: Inconsistent handling of -1.

[Bug tree-optimization/19788] New: Inconsistent handling of -1.

[kazu@cs.umass.edu]

Consider:

int
foo (void)
{
  return -1;
}

int
bar (void)
{
  unsigned int a = 0xffffffff;
  return a;
}

int
baz (void)
{
  int a = -1;
  unsigned int b = 0xffffffff;
  return a == (int) b;
}

Here is what I get:

;; Function foo (foo)

foo ()
{
<bb 0>:
  return -1;

}

;; Function bar (bar)

bar ()
{
  int D.1124;

<bb 0>:
  D.1124 = -000000001;
  return D.1124;

}

;; Function baz (baz)

baz ()
{
  int b.0;

<bb 0>:
  b.0 = -000000001;
  return b.0 == -1;

}

Note that -1 is represented in two different ways in foo and bar.
Worse yet, the tree optimizers do not know that they are equal in baz.
baz is optimized as expected at RTL level.

-- 
           Summary: Inconsistent handling of -1.
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: missed-optimization, TREE
          Severity: enhancement
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: kazu at cs dot umass dot edu
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[kazu@cs.umass.edu]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |19721
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[pinskia@gcc.gnu.org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-06 16:04 -------
Confirmed.  The problem is from TREE_OVERFLOW/TREE_CONSTANT_OVERFLOW.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-02-06 16:04:44
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[schlie@comcast.net]

------- Additional Comments From schlie at comcast dot net  2005-02-06 17:14 -------
(In reply to comment #1)

- as I'm curious as to what the "proper" interpretation/handling of
  target dependant constant value casts should be; it seems that in
  the provided example, the optimized transformations of 0xffffffff
  to an (int)-1 intermediate value in bar & baz are consistent, assuming:

  (((0xffffffff) >> ((sizeof(int) * BITS_PER_UNIT)) & 1)

  however would have expected the comparison between a & b in baz
  to have been optimized away; is this a reasonable expectation?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[pinskia@gcc.gnu.org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-06 17:17 -------
(In reply to comment #2)
> (In reply to comment #1)
Really TREE_OVERFLOW/TREE_CONSTANT_OVERFLOW should go away, there has been some talk about 
removing them.  on the tree level, having that flag set we really say the constant is no longer be able to 
prograte which causes the real problem.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-07 20:17 -------
(In reply to comment #3)
> (In reply to comment #2)
> > (In reply to comment #1)
> Really TREE_OVERFLOW/TREE_CONSTANT_OVERFLOW should go away, there has been some talk 
about 
> removing them.  on the tree level, having that flag set we really say the constant is no longer be able 
to 
> prograte which causes the real problem.

Note I am talking about:
http://www.srcf.ucam.org/~jsm28/gcc/

"Similarly, it should also be possible to get rid of at least one of TREE_OVERFLOW and 
TREE_CONSTANT_OVERFLOW, which also describe what are essentially front-end properties."

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-03-16 12:30 -------
Someone should check if this is fixed now, by this patch:
http://gcc.gnu.org/ml/gcc-patches/2005-03/msg01122.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |stevenb at suse dot de
             Status|NEW                         |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[kazu at cs dot umass dot edu]

------- Additional Comments From kazu at cs dot umass dot edu  2005-03-16 14:53 -------
Here is what I get from the current mainline.

;; Function foo (foo)

foo ()
{
  int D.1132;

<bb 0>:
  return -1;

}



;; Function bar (bar)

bar ()
{
  unsigned int a;
  int D.1137;

<bb 0>:
  return -000000001;

}



;; Function baz (baz)

baz ()
{
  unsigned int b;
  int a;
  int b.0;
  int D.1143;

<bb 0>:
  return 1;

}

So the problem seems to be fixed.
It would still be nice to see -1 instead of -000000001 though.


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788

[Bug tree-optimization/19788] Inconsistent handling of -1.

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19788