[Bug c/25985] New: with optimization integer math fails

[Bug c/25985] New: with optimization integer math fails

[jurka at ejurka dot com]

gcc version 4.2.0 20060121

With optimzation the following code skips the last loop iteration:

#include <stdio.h>

int main(void)
{
        int bits = 25;
        while (bits) {
                printf("bits=%d\n",bits);
                if (bits >= 8) {
                        bits -= 8;
                } else {
                        bits = 0;
                }

        }
        return 0;
}

jurka@fred:~/pg/test$ ~/gcc/42/bin/gcc  ip2.c 
jurka@fred:~/pg/test$ ./a.out 
bits=25
bits=17
bits=9
bits=1
jurka@fred:~/pg/test$ ~/gcc/42/bin/gcc -O ip2.c 
jurka@fred:~/pg/test$ ./a.out 
bits=25
bits=17
bits=9
jurka@fred:~/pg/test$


-- 
           Summary: with optimization integer math fails
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jurka at ejurka dot com
 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=25985

[Bug middle-end/25985] [4.2 Regression] with optimization integer math fails

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-01-26 23:28 -------
Confirmed, quickly looking at this DOM gets it wrong.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2006-01-26 23:28:04
               date|                            |
            Summary|with optimization integer   |[4.2 Regression] with
                   |math fails                  |optimization integer math
                   |                            |fails
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |blocker


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-03-15 18:27 -------
Loop 1 iterates 2 times.
Added canonical iv to loop 1, 2 iterations.

That is wrong, it iterates 3 times at this point (an interation has already
been peeled before the loop).

This is from .ivcanon.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2006-03-15 18:40 -------
Hmm, if I change the function to be:
#include <stdio.h>

int main(void)
{
        int bits = 25;
        while (bits) {
                printf("bits=%d\n",bits);
                if (bits >= 8) {
                        bits -= 8;
                } else {
                        break;
                }

        }
        return 0;
}

And then compare 4.1 vs 4.2, I get the following observation:
in 4.1, it changed bits > 7 to use the iv, while in 4.2, it changed the bits !=
0 to use the iv


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2006-03-15 20:15 -------
I'm sure Zdenek has an idea...


-- 

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=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rakdver at gcc dot gnu dot org]

-- 

rakdver at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rakdver at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2006-01-26 23:28:04         |2006-03-15 20:22:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rakdver at gcc dot gnu dot org]

------- Comment #5 from rakdver at gcc dot gnu dot org  2006-03-16 02:49 -------
"Added canonical iv..." seems correct to me (at least, -fno-tree-loop-ivcanon
does not fix the problem).  However,  -fno-tree-loop-ivcanon -fno-ivopts fixes
it, so it is mine anyway.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rakdver at gcc dot gnu dot org]

------- Comment #6 from rakdver at gcc dot gnu dot org  2006-03-16 03:15 -------
Actually, my previous comment is wrong, it is ivcanon (more precisely, # of
iterations analysis).  Consider this testcase:

#include <stdio.h>

void bla(int x)
{
  if (x < -100)
   exit (0);
}

int main(void)
{
  int bits = 25;
  while (bits) {
      printf ("%d\n", bits);
      bla(bits);
      bits -= 8;
  }
  return 0;
}

We note that bits is signed, therefore it cannot overflow, and conclude the
loop must be finite.  Using this assumption, we conclude it iterates two times.
However, this would only be correct if the loop were exited through this
particular exit, as these examples demonstrate.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rakdver at gcc dot gnu dot org]

------- Comment #7 from rakdver at gcc dot gnu dot org  2006-03-16 14:29 -------
Patch:

http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01019.html


-- 

rakdver at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2006-
                   |                            |03/msg01019.html
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[rakdver at gcc dot gnu dot org]

------- Comment #8 from rakdver at gcc dot gnu dot org  2006-03-29 01:41 -------
Subject: Bug 25985

Author: rakdver
Date: Wed Mar 29 01:41:27 2006
New Revision: 112484

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=112484
Log:
        PR tree-optimization/25985
        * tree-ssa-loop-niter.c (number_of_iterations_le,
        number_of_iterations_ne): Make comments more precise.
        (number_of_iterations_cond): Add only_exit argument.  Use the
        fact that signed variables do not overflow only when only_exit
        is true.
        (loop_only_exit_p): New.
        (number_of_iterations_exit): Pass result of loop_only_exit_p to
        number_of_iterations_cond.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-ssa-loop-niter.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2006-04-11 23:13 -------
Fixed.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985

[Bug tree-optimization/25985] [4.2 Regression] with optimization integer math fails

[jakub at gcc dot gnu dot org]

------- Comment #10 from jakub at gcc dot gnu dot org  2006-05-04 06:40 -------
Subject: Bug 25985

Author: jakub
Date: Thu May  4 06:40:15 2006
New Revision: 113515

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=113515
Log:
        PR tree-optimization/27285

        Backport from mainline:

        2006-03-28  Zdenek Dvorak  <dvorakz@suse.cz>

        PR tree-optimization/25985
        * tree-ssa-loop-niter.c (number_of_iterations_le,
        number_of_iterations_ne): Make comments more precise.
        (number_of_iterations_cond): Add only_exit argument.  Use the
        fact that signed variables do not overflow only when only_exit
        is true.
        (loop_only_exit_p): New.
        (number_of_iterations_exit): Pass result of loop_only_exit_p to
        number_of_iterations_cond.

        * gcc.c-torture/execute/pr27285.c: New test.

Added:
    branches/gcc-4_1-branch/gcc/testsuite/gcc.c-torture/execute/pr27285.c
Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_1-branch/gcc/tree-ssa-loop-niter.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25985