[Bug other/17100] New: Missed opportunity for value range optimization

[Bug other/17100] New: Missed opportunity for value range optimization

[gcc-bugzilla at gcc dot gnu dot org]

Description:
A non-optimal code sequence is illustraded.    Duplicate using gcc 3.5 and
command line:

gcc -O3 -m64 -c test.c


Testcase:
int x;
int a[10];

void bar () {
  printf("bar\n");
}

void foo ()
{
  int i;
  int j;

  j = x;

  for (i = 0; i < 10; i++)
  {
    j += a[i];

    /* This block is successfully removed.  GCC figures
     out that 15 is out of the range of i (0-9) and
     removes the block */
    if (i == 15) bar ();

    /* This block is not removed, but could be */
    if (i == -5) bar ();
  }

}

Assembly:
.foo:
      mflr 0
      std 31,-8(1)
      li 31,0
      std 0,16(1)
      stdu 1,-128(1)
.L12:
      addi 0,31,1
      extsw 31,0
      cmpwi 7,31,9
      cmpwi 6,31,-5
      bgt- 7,.L14
.L4:
      bne+ 6,.L12
      ld 3,.LC2@toc(2)
      bl .puts
      nop
      addi 0,31,1
      extsw 31,0
      cmpwi 7,31,9
      cmpwi 6,31,-5
      ble+ 7,.L4
.L14:
      addi 1,1,128
      ld 0,16(1)
      ld 31,-8(1)
      mtlr 0
      blr



-- 
           Summary: Missed opportunity for value range optimization
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P1
         Component: other
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: steinmtz at us dot ibm dot com
                CC: gcc-bugs at gcc dot gnu dot org,steinmtz at us dot ibm
                    dot com
 GCC build triplet: powerpc64-linux
  GCC host triplet: powerpc64-linux
GCC target triplet: powerpc64-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-08-19 15:03 -------
Confirmed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|other                       |tree-optimization
     Ever Confirmed|                            |1
           Keywords|                            |missed-optimization
   Last reconfirmed|0000-00-00 00:00:00         |2004-08-19 15:03:47
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[nathan at gcc dot gnu dot org]

------- Additional Comments From nathan at gcc dot gnu dot org  2004-10-12 10:34 -------
I think this is an artifact of the design of the current value range
propagation.  That's not really range propagation by comparison result CSEing. 
The loop exit is i < 10, and the next comparison is i == 15.  We can propagate
the 'i < 10' result
and deduce that i == 15 must be false.  The later i == -5 can't be determined
from the result of i < 10.

I suspect this will be much easier to fix with a proper SSA VRP pass.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[sebastian dot pop at cri dot ensmp dot fr]

------- Additional Comments From sebastian dot pop at cri dot ensmp dot fr  2004-12-19 23:43 -------
Patch in autovect-branch:
http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01444.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-04-24 11:48 -------
i_8: VARYING


  # i_8 = PHI <0(0), i_10(4)>;
<L2>:;
  if (i_8 == -5) goto <L3>; else goto <L8>;


  i_10 = i_8 + 1;
  if (i_10 <= 9) goto <L2>; else goto <L6>;

i_10: [1, 2147483647]

Hmm, isn't i_10 signed still so why between 1 and -1, that does not make sense.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[dnovillo at gcc dot gnu dot org]

------- Additional Comments From dnovillo at gcc dot gnu dot org  2005-06-02 18:50 -------

Fixed.  Likely with http://gcc.gnu.org/ml/gcc-patches/2005-06/msg00127.html, but
it may have been fixed earlier.  The natural range for i_6 is computed as [0, 9]
so this isn't something that was folded using range equivalences.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100

[Bug tree-optimization/17100] Missed opportunity for value range optimization

[pinskia at gcc dot gnu dot org]

-- 
Bug 17100 depends on bug 18373, which changed state.

Bug 18373 Summary: [meta-bug] VRP Value Range Propagation
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18373

           What    |Old Value                   |New Value
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17100