[Bug tree-optimization/36188] New: missed CCP

[Bug tree-optimization/36188] New: missed CCP

[rguenth at gcc dot gnu dot org]

int f(int lay)
{
  static int syncsize, init;
  if (init == 0)
     syncsize = 1344, init = 1;
  return syncsize;
}


store-ccp should handle that.


-- 
           Summary: missed CCP
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          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=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2008-05-09 15:35 -------
A get_symbol_constant_value at the right place and time should do the trick.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2008-05-09 15:39 -------
I thought I had a bug filed already ...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2008-05-13 14:43 -------
So the problem is that we are not optimistically treating static variables.  If
we fix that then the testcase is optimized to return 1344 with store_ccp.  The
question is how many bugs we hit with that and how restrictive we need to be ;)


-- 

rguenth 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         |2008-05-13 14:43:14
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2008-05-13 14:44 -------
Created an attachment (id=15635)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15635&action=view)
very optimistic patch

Which also handles

int f(int lay)
{
  static int syncsize, init;
  if (init == 1)
     syncsize = 1344, init = 1;
  return syncsize;
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #5 from rguenth at gcc dot gnu dot org  2008-05-13 15:01 -------
Ok, that doesn't really work well.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2008-08-15 23:52 -------
On the trunk we get:
        lis 11,.LANCHOR0@ha
        la 9,.LANCHOR0@l(11)
        lwz 0,.LANCHOR0@l(11)
        cmpwi 7,0,0
        bne 7,.L6
        li 0,1344
        li 3,1344
        stw 0,4(9)
        li 9,1
        stw 9,.LANCHOR0@l(11)
        blr
Which is better but still can be improved.  Basically we have to prove syncsize
is only set when init is set.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2009-04-22 21:56 -------
This also happens in 189.lucas where it prevents propagating constant 1 array
strides to the access.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[steven at gcc dot gnu dot org]

------- Comment #8 from steven at gcc dot gnu dot org  2009-04-25 22:24 -------
Re. comment #5 -- what doesn't work very well, i.e. what massive breakage does
your patch cause?

Maybe you can treat static locals optimistically if they are only stored to
once?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188

[Bug tree-optimization/36188] missed CCP

[rguenth at gcc dot gnu dot org]

------- Comment #9 from rguenth at gcc dot gnu dot org  2009-04-26 16:36 -------
comment #5 doesn't work anymore because CCP no longer tracks stores.

What we would need is predication in value-numbering.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dberlin at gcc dot gnu dot
                   |                            |org, rguenth at gcc dot gnu
                   |                            |dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36188