[Bug tree-optimization/48739] New: ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[Bug tree-optimization/48739] New: ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[arthur.j.odwyer at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

           Summary: ICE in check_loop_closed_ssa_use() with
                    "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: arthur.j.odwyer@gmail.com


Created attachment 24077
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24077
Output of "ajo-gcc -w -O1 -ftree-parallelize-loops=2 -fno-tree-dominator-opts
-c test1942241936.c -v"

This reproduces for me with svn revision 172796 (2011-04-20), but it doesn't
reproduce with gcc-4.5.  I'm on Ubuntu 10.10, x86-64.  Attached "gcc-v.txt".

cat >test1942241936.c <<EOF
extern int g;
extern void func_48(void);
int func_70(int p) {
    int a, b, *ptr = &g;
    for (a = 0; a < 10; ++a) {
        func_48();
        for (b = 0; b < 5; ++b) {
            for (g = 0; g < 1; ++g) {
                p = 0;
                ptr = &p;
            }
        }
    }
    *ptr = p;
    for (p = 0; p != 10; p += 3u) { }
    return g;
}
EOF
gcc -w -O1 -ftree-parallelize-loops=2 -fno-tree-dominator-opts -c
test1942241936.c

test1942241936.c: In function ‘func_70’:
test1942241936.c:3:5: internal compiler error: in check_loop_closed_ssa_use, at
tree-ssa-loop-manip.c:422


This test case is reduced from the output of Csmith
(http://embed.cs.utah.edu/csmith/), using the following command line:
csmith --bitfields --packed-struct -s 1942241936 >test1942241936.c

Notice that the final for-loop has undefined behavior due to signed integer
overflow, which might be putting the compiler in an unfamiliar state... but
surely it shouldn't *crash* the compiler! -fwrapv doesn't make the crash go
away, either.

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org
   Target Milestone|---                         |4.5.4
            Summary|ICE in                      |[4.5/4.6/4.7 Regression]
                   |check_loop_closed_ssa_use() |ICE in
                   |with                        |check_loop_closed_ssa_use()
                   |"-ftree-parallelize-loops=2 |with
                   |-fno-tree-dominator-opts"   |"-ftree-parallelize-loops=2
                   |                            |-fno-tree-dominator-opts"

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-19 16:38:45 UTC ---
Simplified testcase:

/* PR tree-optimization/48739 */
/* { dg-do compile } */
/* { dg-options "-O1 -ftree-parallelize-loops=2 -fno-tree-dominator-opts" } */

extern int g;
extern void bar (void);

int
foo (int x)
{
  int a, b, *c = (int *) 0;
  for (a = 0; a < 10; ++a)
    {
      bar ();
      for (b = 0; b < 5; ++b)
        {
          x = 0;
          c = &x;
          g = 1;
        }
    }
  *c = x;
  for (x = 0; x != 10; x++)
    ;
  return g;
}

Started with http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=149206
(before that the empty loop would be just removed).

The bug seems to be that parloops pass requests TODO_rebuild_alias which
changes x from TREE_ADDRESSABLE to a gimple var, but we are in loop closed SSA
form at that point and nothing ensures that the optimized variable is put into
loop closed SSA form.

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-08-19
     Ever Confirmed|0                           |1

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-19 16:43:03 UTC ---
If the loop is non-empty, such as containing v[x] = x; and extern int v[10];,
then it started failing with
http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=145494

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-19 16:59:00 UTC ---
Created attachment 25055
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25055
gcc47-pr48739.patch

Untested fix.

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-08-19 23:04:36 UTC ---
(In reply to comment #1)
> Simplified testcase:
> 
> /* PR tree-optimization/48739 */
> /* { dg-do compile } */
> /* { dg-options "-O1 -ftree-parallelize-loops=2 -fno-tree-dominator-opts" } */
> 
> extern int g;
> extern void bar (void);
> 
> int
> foo (int x)
> {
>   int a, b, *c = (int *) 0;
>   for (a = 0; a < 10; ++a)
>     {
>       bar ();
>       for (b = 0; b < 5; ++b)
>         {
>           x = 0;
>           c = &x;
>           g = 1;
>         }
>     }
>   *c = x;
>   for (x = 0; x != 10; x++)
>     ;
>   return g;
> }
> 
> Started with http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=149206
> (before that the empty loop would be just removed).
> 
> The bug seems to be that parloops pass requests TODO_rebuild_alias which
> changes x from TREE_ADDRESSABLE to a gimple var, but we are in loop closed SSA
> form at that point and nothing ensures that the optimized variable is put into
> loop closed SSA form.

The bug would be that TODO_rebuild_alias does a TODO_update_address_taken.
But really parloops should be re-architected to not require a
TODO_rebuild_alias ... (it isn't required anymore for correctness anyway)

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-20 07:48:41 UTC ---
Author: jakub
Date: Sat Aug 20 07:48:35 2011
New Revision: 177924

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=177924
Log:
    PR tree-optimization/48739
    * tree-ssa.c: Include cfgloop.h.
    (execute_update_addresses_taken): When updating ssa, if in
    loop closed SSA form, call rewrite_into_loop_closed_ssa instead of
    update_ssa.
    * Makefile.in (tree-ssa.o): Depend on $(CFGLOOP_H).

    * gcc.dg/pr48739-1.c: New test.
    * gcc.dg/pr48739-2.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr48739-1.c
    trunk/gcc/testsuite/gcc.dg/pr48739-2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/Makefile.in
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa.c

[Bug tree-optimization/48739] [4.5/4.6/4.7 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-20 07:51:12 UTC ---
Author: jakub
Date: Sat Aug 20 07:51:09 2011
New Revision: 177925

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=177925
Log:
    PR tree-optimization/48739
    * tree-ssa.c: Include cfgloop.h.
    (execute_update_addresses_taken): When updating ssa, if in
    loop closed SSA form, call rewrite_into_loop_closed_ssa instead of
    update_ssa.
    * Makefile.in (tree-ssa.o): Depend on $(CFGLOOP_H).

    * gcc.dg/pr48739-1.c: New test.
    * gcc.dg/pr48739-2.c: New test.

Added:
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/pr48739-1.c
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/pr48739-2.c
Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/Makefile.in
    branches/gcc-4_6-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_6-branch/gcc/tree-ssa.c

[Bug tree-optimization/48739] [4.5 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.5/4.6/4.7 Regression]    |[4.5 Regression] ICE in
                   |ICE in                      |check_loop_closed_ssa_use()
                   |check_loop_closed_ssa_use() |with
                   |with                        |"-ftree-parallelize-loops=2
                   |"-ftree-parallelize-loops=2 |-fno-tree-dominator-opts"
                   |-fno-tree-dominator-opts"   |

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-20 07:57:04 UTC ---
Fixed for 4.6+ so far.

[Bug tree-optimization/48739] [4.5 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug tree-optimization/48739] [4.5 Regression] ICE in check_loop_closed_ssa_use() with "-ftree-parallelize-loops=2 -fno-tree-dominator-opts"

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48739

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|4.5.4                       |4.6.2

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-07-02 09:51:43 UTC ---
Fixed for 4.6.2.