[Bug tree-optimization/28386] New: [4.1 regression] Wrong code with if condition in loop

[Bug tree-optimization/28386] New: [4.1 regression] Wrong code with if condition in loop

[reichelt at gcc dot gnu dot org]

Consider the following testcase - a variant from
http://gcc.gnu.org/ml/gcc-help/2006-07/msg00177.html :

=========================================================
#include <stdio.h>

volatile char s[256][3];

int main()
{
    int i, j=0;

    for (i = 0; i < 256; i++)
        if (i >= 128 && i < 256)
            { printf("%d ", s[i - 128][0]); ++j; }
    printf("\nj = %d\n", j);
    return 0;
}
=========================================================

This program should generate the following output
(wrapped to 80 characters per line):

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 
j = 128

With -Os I get the output

-125 4 -21 -8 -16 93 -119 -24 0 -127 17 80 -2 89 -61 3 0 2 100 10 61 100 0 0 0
0
 -107 0 0 0 0 12 0 4 0 -108 8 0 -127 5 0 4 0 112 8 0 0 11 0 0 0 -28 64 0 -107 2 
0 0 0 17 0 0 -126 17 0 4 0 8 0 0 0 -2 111 4 -1 1 0 -1 -126 0 0 0 0 0 0 0 0 0 0
0
 0 0 0 0 0 -1 -1 0 -1 0 0 0 -108 -8 64 0 -52 -80 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 
j = 255

Parts of the if-condition ionside the loop have been optimized out.
Happens only on the 4.1 branch.


-- 
           Summary: [4.1 regression] Wrong code with if condition in loop
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: wrong-code, monitored
          Severity: blocker
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: reichelt at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug tree-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[reichelt at gcc dot gnu dot org]

------- Comment #1 from reichelt at gcc dot gnu dot org  2006-07-14 20:10 -------
Confirmed.


-- 

reichelt 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         |2006-07-14 20:10:03
               date|                            |
   Target Milestone|---                         |4.1.2
            Version|unknown                     |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug rtl-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2006-07-15 11:40 -------
This is going wrong in RTL land.  Tree optimization looks ok - we're merging
the
two IVs and using a wrapping unsigned IV going from 0xffffff81 to 127:

<bb 0>:
  ivtmp.33D.1833 = 0ffffff81;
  jD.1769 = 0;

<L0>:;
  if (ivtmp.33D.1833 <= 127) goto <L1>; else goto <L2>;

<L1>:;
  D.1775 = (intD.0) ivtmp.33D.1833;
  D.1776 = sD.1765[D.1775][0];
  printf (&"%d "[0], (intD.0) D.1776);
  jD.1769 = jD.1769 + 1;

<L2>:;
  ivtmp.33D.1833 = ivtmp.33D.1833 + 1;
  if (ivtmp.33D.1833 != 128) goto <L0>; else goto <L4>;

<L4>:;
  printf (&"\nj = %d\n"[0], jD.1769);
  return 0;

of course, it's the particular choice of IVs that triggers the bug later on.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
          Component|tree-optimization           |rtl-optimization


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug rtl-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2006-07-15 11:54 -------
Goes away with -floop-optimize2.  *sigh*


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug rtl-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[ebotcazou at gcc dot gnu dot org]

------- Comment #4 from ebotcazou at gcc dot gnu dot org  2006-08-11 07:31 -------
Investigating.


-- 

ebotcazou at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |ebotcazou at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2006-07-14 20:10:03         |2006-08-11 07:31:09
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug rtl-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[ebotcazou at gcc dot gnu dot org]

------- Comment #5 from ebotcazou at gcc dot gnu dot org  2006-09-05 07:06 -------
Subject: Bug 28386

Author: ebotcazou
Date: Tue Sep  5 07:06:46 2006
New Revision: 116693

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=116693
Log:
        PR rtl-optimization/28386
        * loop.c (biased_biv_may_wrap_p): Rename to biv_may_wrap_p and
        remove 'bias' parameter.
        (maybe_eliminate_biv_1): Adjust for above change.


Added:
    branches/gcc-4_1-branch/gcc/testsuite/gcc.c-torture/execute/20060905-1.c
Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/loop.c
    branches/gcc-4_1-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386

[Bug rtl-optimization/28386] [4.1 regression] Wrong code with if condition in loop

[ebotcazou at gcc dot gnu dot org]

------- Comment #6 from ebotcazou at gcc dot gnu dot org  2006-09-05 07:08 -------
Fixed in upcoming 4.1.2 release.


-- 

ebotcazou at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28386