[Bug c/42006] New: Termination problem with -O2 and -O3

Using r154087, the code below never terminates with -O2 and -O3.

yang@yang-working:~$ svngcc -O2 -o small2 small.c
yang@yang-working:~$ ./small2
1
1
...
yang@yang-working:~$ ./small1
1
2
exit!
yang@yang-working:~$ cat small.c
#include <stdio.h>

static int
my_add(int si1, int si2)
{
    return (si1 > (50-si2)) ? si1 : (si1 + si2);
}

static unsigned int
my_shift(unsigned int left, int right)
{
  return  (right > 100) ? left : (left >> right);
}

static int func_4(unsigned int p_6)
{
    for (p_6 = 1; p_6 < 3; p_6 = my_add(p_6, 1))
    {
            printf ("%d\n", p_6);
            if (my_shift(p_6, p_6))
            {
                return 0;
            }
    }
    return 0;
}

int main(void)
{
    func_4(0);
    printf("exit!\n");
    return 0;
}

yang@yang-working:~$ svngcc -v
Using built-in specs.
COLLECT_GCC=svngcc
COLLECT_LTO_WRAPPER=/home/yang/compilers/libexec/gcc/i686-pc-linux-gnu/4.5.0/lto-wrapper
Target: i686-pc-linux-gnu
Configured with: ../configure --enable-lto --prefix=/home/yang/compilers
--program-prefix=svn --enable-languages=c,c++
--with-libelf=/home/yang/compilers : (reconfigured) ../configure --enable-lto
--prefix=/home/yang/compilers --program-prefix=svn
--with-libelf=/home/yang/compilers --enable-languages=c,lto,c++ --no-create
--no-recursion : (reconfigured) ../configure --enable-lto
--prefix=/home/yang/compilers --program-prefix=svn
--with-libelf=/home/yang/compilers --enable-languages=c,c++,lto --no-create
--no-recursion : (reconfigured) ../configure --enable-lto
--prefix=/home/yang/compilers --program-prefix=svn
--with-libelf=/home/yang/compilers --enable-languages=c,c++,lto --no-create
--no-recursion : (reconfigured) ../configure --enable-lto
--prefix=/home/yang/compilers --program-prefix=svn
--with-libelf=/home/yang/compilers --enable-languages=c,c++,lto --no-create
--no-recursion
Thread model: posix
gcc version 4.5.0 20091110 (experimental) (GCC)


--
           Summary: Termination problem with -O2 and -O3
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: chenyang at cs dot utah dot edu
 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=42006

------- Comment #1 from rguenth at gcc dot gnu dot org  2009-11-11 20:35 -------
Confirmed.


--

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
           Keywords|                            |wrong-code
      Known to work|                            |4.4.2
   Last reconfirmed|0000-00-00 00:00:00         |2009-11-11 20:35:30
               date|                            |
            Summary|Termination problem with -O2|[4.5 Regression] Termination
                   |and -O3                     |problem with -O2 and -O3
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006

------- Comment #2 from rguenth at gcc dot gnu dot org  2009-11-27 11:17 -------
Re-confirmed.


--

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |middle-end
  GCC build triplet|i686-pc-linux-gnu           |
   GCC host triplet|i686-pc-linux-gnu           |
 GCC target triplet|i686-pc-linux-gnu           |x86_64-*-*, i?86-*-*
           Priority|P3                          |P1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006

------- Comment #3 from rguenth at gcc dot gnu dot org  2009-11-27 14:18 -------
This is early IPA-SRA which transforms

<bb 2>:
  goto <bb 9>;

<bb 3>:
  p_6.0_3 = (int) p_6_1;
  if (p_6.0_3 <= 100)
    goto <bb 4>;
  else
    goto <bb 5>;

<bb 4>:
  p_6_6 = p_6_1 >> p_6.0_3;

<bb 5>:
  # p_6_7 = PHI <p_6_1(3), p_6_6(4)>
  if (p_6_7 != 0)
    goto <bb 10>;
  else
    goto <bb 6>;

<bb 6>:
  if (p_6_1 <= 49)
    goto <bb 7>;
  else
    goto <bb 8>;

<bb 7>:
  p_6_9 = p_6_1 + 1;

<bb 8>:
  # p_6_10 = PHI <p_6_1(6), p_6_9(7)>

<bb 9>:
  # p_6_1 = PHI <1(2), p_6_10(8)>
  if (p_6_1 <= 2)
    goto <bb 3>;
  else
    goto <bb 10>;

to

<bb 2>:
  goto <bb 9>;

<bb 3>:
  p_6.0_3 = (int) p_6_2;
  if (p_6.0_3 <= 100)
    goto <bb 4>;
  else
    goto <bb 5>;

<bb 4>:
  p_6_5 = p_6_2 >> p_6.0_3;

<bb 5>:
  # p_6_12 = PHI <p_6_2(3), p_6_5(4)>
  if (p_6_12 != 0)
    goto <bb 10>;
  else
    goto <bb 6>;

<bb 6>:
  if (p_6_12 <= 49)
    goto <bb 7>;
  else
    goto <bb 8>;

<bb 7>:
  p_6_11 = p_6_12 + 1;

<bb 8>:
  # p_6_4 = PHI <p_6_12(6), p_6_11(7)>

<bb 9>:
  # p_6_2 = PHI <1(2), p_6_4(8)>
  if (p_6_2 <= 2)
    goto <bb 3>;
  else
    goto <bb 10>;


see how p_6_9 = p_6_1 + 1 is changed to p_6_11 = p_6_12 + 1 forming an
endless loop - it shuld have been p_6_11 = p_6_2 + 1 instead.

Martin?


--

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006

------- Comment #4 from jamborm at gcc dot gnu dot org  2009-11-27 18:46 -------
I have started a bootstrap and check of the following (with a testcase
and change log and stuff...) which for some reason fixes this
(renaming introduces the mess, the code as is generated by IPA-SRA is OK):

Index: mine/gcc/tree-sra.c
===================================================================
--- mine.orig/gcc/tree-sra.c
+++ mine/gcc/tree-sra.c
@@ -3478,7 +3478,10 @@ get_replaced_param_substitute (struct ip
     {
       char *pretty_name = make_fancy_name (adj->base);

-      repl = make_rename_temp (TREE_TYPE (adj->base), "ISR");
+      repl = create_tmp_var (TREE_TYPE (adj->base), "ISR");
+      if (TREE_CODE (TREE_TYPE (repl)) == COMPLEX_TYPE
+         || TREE_CODE (TREE_TYPE (repl)) == VECTOR_TYPE)
+       DECL_GIMPLE_REG_P (repl) = 1;
       DECL_NAME (repl) = get_identifier (pretty_name);
       obstack_free (&name_obstack, pretty_name);


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006

------- Comment #5 from jamborm at gcc dot gnu dot org  2009-11-27 22:34 -------
Subject: Bug 42006

Author: jamborm
Date: Fri Nov 27 22:34:40 2009
New Revision: 154715

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154715
Log:
2009-11-27  Martin Jambor  <mjambor@...>

        PR middle-end/42006
        * tree-sra.c (get_replaced_param_substitute): Call create_tmp_var
        instead of create_tmp_var.  Set DECL_GIMPLE_REG_P to one manually
        for vector and complex types.
        (get_adjustment_for_base): Describe return value in the comment.

        * testsuite/gcc.c-torture/execute/pr42006.c: New test.


Added:
    trunk/gcc/testsuite/gcc.c-torture/execute/pr42006.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-sra.c


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006

------- Comment #6 from jamborm at gcc dot gnu dot org  2009-11-27 22:35 -------
Fixed.


--

jamborm at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42006