[Bug tree-optimization/19581] New: Missed store monition on the tree level (on PPC in general)

[Bug tree-optimization/19581] New: Missed store monition on the tree level (on PPC in general)

[pinskia at gcc dot gnu dot org]

This is like PR 19580 but a request to optimize it on the tree level instead of just on the RTL level.  
Mainly because we don't catch it for PPC (with 3.3.3 or with -fno-gcse).
int r[6];

void f (int n)
{
  while (-- n)
    {
      r [0] += r [5];
      r [1] += r [0];
      r [2] += r [1];
      r [3] += r [2];
      r [4] += r [3];
      r [5] += r [4];
    }
}

We should be able to pull out the load and stores for r[0] ... r[5].
If we do it on the tree level, we also just fix the problem for 4.0 also (but I think this might need more 
aliasing analysis or I could be wrong).

-- 
           Summary: Missed store monition on the tree level (on PPC in
                    general)
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, alias, TREE
          Severity: enhancement
          Priority: P2
         Component: tree-optimization
        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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level (on PPC in general)

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |19580
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level (on PPC in general)

[dberlin at gcc dot gnu dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-01-23 01:47 -------
This needs more alias analysis
Structure aliasing could do this with a little work, so i'll take this for now

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |dberlin at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-01-23 01:47:51
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level (on PPC in general)

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-01-23 19:27 -------
Right now it is the *old* loop optimizer that promotes the array elements 
to registers.  How about that. 
 
Is there an "old loop optimizer" meta bug?  We'll probably want to do this 
elsewhere before the old loop optimizer can die. 
 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rakdver at gcc dot gnu dot
                   |                            |org, steven at gcc dot gnu
                   |                            |dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-01-23 19:29 -------
Has nothing to do with PPC 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Missed store monition on the|Missed store monition on the
                   |tree level (on PPC in       |tree level
                   |general)                    |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level

[rakdver at gcc dot gnu dot org]

------- Additional Comments From rakdver at gcc dot gnu dot org  2005-01-23 19:44 -------
Gcse store motion could catch this as well (if I understand PR 19580 well, it
does not, which is a bug), so this pr alone would not be a reason to block
removal of old loop optimizer.  However hopefully we should have a usable alias
analysis
on tree level in 4.1, of course.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-01-23 19:53 -------
GCSE store motion does catch this.  But it is disabled for GCC 4.0 because 
it is buggy and does not really work well in most cases. 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-23 19:59 -------
(In reply to comment #5)
> GCSE store motion does catch this.  But it is disabled for GCC 4.0 because 
> it is buggy and does not really work well in most cases. 
Yes but we don't catch in 3.4.0 so that means GCSE store motion did not catch it since GCSE store 
motion was only disabled for 4.0.0 but was enabled after the rewrite for 3.4.0.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store monition on the tree level

[rakdver at gcc dot gnu dot org]

------- Additional Comments From rakdver at gcc dot gnu dot org  2005-01-23 20:01 -------
The same holds for the old loop optimizer :-)

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store motion on the tree level

[dberlin at gcc dot gnu dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-26 13:18 -------
Richard guenther is working on aliasing for arrays, i'll leave this one for him

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
         AssignedTo|dberlin at gcc dot gnu dot  |unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store motion on the tree level

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-07-26 13:29 -------
For a reduced array with only 4 elements (I know - this should be a --param) we
now get in .vars (with the array aliasing patch):

f (n)
{
  int n.39;
  unsigned int ivtmp.33;
  int lsm_tmp.32;
  int lsm_tmp.31;
  int lsm_tmp.30;
  int lsm_tmp.29;

<bb 0>:
  n.39 = n - 1;
  if (n.39 != 0) goto <L6>; else goto <L2>;

<L6>:;
  lsm_tmp.29 = r[2];
  lsm_tmp.30 = r[1];
  lsm_tmp.31 = r[3];
  lsm_tmp.32 = r[0];
  ivtmp.33 = 0;

<L0>:;
  lsm_tmp.32 = lsm_tmp.32 + lsm_tmp.31;
  lsm_tmp.30 = lsm_tmp.32 + lsm_tmp.30;
  lsm_tmp.29 = lsm_tmp.30 + lsm_tmp.29;
  lsm_tmp.31 = lsm_tmp.31 + lsm_tmp.29;
  ivtmp.33 = ivtmp.33 + 1;
  if (ivtmp.33 != (unsigned int) n.39) goto <L0>; else goto <L10>;

<L10>:;
  r[2] = lsm_tmp.29;
  r[1] = lsm_tmp.30;
  r[3] = lsm_tmp.31;
  r[0] = lsm_tmp.32;

<L2>:;
  return;

}

and the asm loop looks like

.L4:
        addl    %edx, %eax
        incl    %esi
        addl    %eax, %ecx
        addl    %ecx, %ebx
        addl    %ebx, %edx
        cmpl    %edi, %esi
        jne     .L4

as in exactly what you want.  Don't hold your breath for 4.1, though.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2005-06-09 16:37:28         |2005-07-26 13:29:06
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store motion on the tree level

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2006-02-03 12:32 -------
On trunk with --param salias-max-array-elements=6 we now get

.L4:
        addl    %edx, %eax
        addl    %eax, %ecx
        addl    %ecx, %ebx
        addl    %ebx, %esi
        addl    %esi, %edi
        addl    %edi, %edx
        subl    $1, -16(%ebp)
        jne     .L4

so this is fixed.  LSM could also be teached to use the infrastrucure I
invented for the copyprop enhancement.


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581

[Bug tree-optimization/19581] Missed store motion on the tree level

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19581