[Bug middle-end/23294] New: fold does not fold a*2-a to a

[Bug middle-end/23294] New: fold does not fold a*2-a to a

[pinskia at gcc dot gnu dot org]

I noticed this while running a profiledbootstrap and looking through simplify-rtx.c.
Take the following function:
int f(int a)
{
  return a*2-a;
}

We don't optimizate a*2-a to just a on the tree level but do in combine, we do.

-- 
           Summary: fold does not fold a*2-a to a
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, TREE
          Severity: enhancement
          Priority: P2
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
OtherBugsDependingO 19986
             nThis:


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*2-a to a

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-10 00:45 -------
Here is an example which is __not__ folded back and shows that we don't optimize it on the RTL level at 
least for i686:
int f(int a)
{
  return a*6-a;
}
int f1(int a)
{
  return a*5;
}

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-10 00:49 -------
Confirmed.

a*C+a is also not simplified.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-10 00:49:31
               date|                            |
            Summary|fold does not fold a*2-a to |fold does not fold a*C+a to
                   |a                           |a*(C+1) or a*C-a to a*(C-1)


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-08-10 09:23 -------
I believe we can only do so for -fwrapv (but we don't) or for unsigned (which we
also don't do).  I may look at this somewhen in the future.

We also don't canonicalize

int f(int a)
{
  return a + a + a;
}

which I believe we did at some time?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-10 10:26 -------
(In reply to comment #3)
> I believe we can only do so for -fwrapv (but we don't) or for unsigned (which we
> also don't do).  I may look at this somewhen in the future.

Why do you believe that with -fno-wrapv, overflow is undefined so anything can happen.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-08-26 09:34 -------
I have a patch.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2005-08-10 00:49:31         |2005-08-26 09:34:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-08-26 13:50 -------
Patch posted.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2005-
                   |                            |08/msg01555.html


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[rguenth at gcc dot gnu dot org]

------- Comment #8 from rguenth at gcc dot gnu dot org  2005-11-19 11:31 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294

[Bug middle-end/23294] fold does not fold a*C+a to a*(C+1) or a*C-a to a*(C-1)

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2005-11-19 11:29 -------
Subject: Bug 23294

Author: rguenth
Date: Sat Nov 19 11:29:10 2005
New Revision: 107218

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=107218
Log:
2005-11-19  Richard Guenther  <rguenther@suse.de>

        PR middle-end/23294
        * fold-const.c (fold_plusminus_mult_expr): New function.
        (fold_binary): Use to canonicalize PLUS_EXPR and MINUS_EXPR
        cases, remove now unnecessary code.

        * gcc.dg/tree-ssa/pr23294.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr23294.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fold-const.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23294