[Bug c/52996] New: ice in verify_loop_structure, at cfgloop.c:1567

[Bug c/52996] New: ice in verify_loop_structure, at cfgloop.c:1567

[regehr at cs dot utah.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

             Bug #: 52996
           Summary: ice in verify_loop_structure, at cfgloop.c:1567
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: regehr@cs.utah.edu
                CC: chenyang@cs.utah.edu


[regehr@dyson r33]$ current-gcc -O3 -c small.c
small.c: In function 'fn1':
small.c:29:1: error: edge from 18 to 19 should be marked irreducible
 }
 ^
small.c:29:1: error: basic block 19 should be marked irreducible
 }
 ^
small.c:29:1: error: edge from 19 to 11 should be marked irreducible
 }
 ^
small.c:29:1: internal compiler error: in verify_loop_structure, at
cfgloop.c:1567
 }
 ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.


[regehr@dyson r33]$ cat small.c
int a, b, c, d, e, f;
void
fn1 ()
{
    int g, h;
lbl_173:
    if (a)
        c++;
lbl_158:
    e = 0;
    h = 0;
    for (; h > -8; h--)
    {
        g = 0;
        for (; g <= 0; g += 1)
        {
            d = 1;
            for (; d >= 0; d -= 1)
                a = 1;
            if (b)
                break;
            if (f)
                return;
            if (h)
                goto lbl_158;
        }
    }
    goto lbl_173;
}


[regehr@dyson r33]$ current-gcc -v
Using built-in specs.
COLLECT_GCC=current-gcc
COLLECT_LTO_WRAPPER=/uusoc/exports/scratch/regehr/z/compiler-install/gcc-r186455-install/bin/../libexec/gcc/x86_64-unknown-linux-gnu/4.8.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../configure --with-libelf=/usr/local --enable-lto
--prefix=/home/regehr/z/compiler-install/gcc-r186455-install
--program-prefix=r186455- --enable-languages=c,c++
Thread model: posix
gcc version 4.8.0 20120414 (experimental) (GCC)

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
          Component|c                           |middle-end
   Target Milestone|---                         |4.8.0
            Summary|ice in                      |[4.8 Regression] ice in
                   |verify_loop_structure, at   |verify_loop_structure, at
                   |cfgloop.c:1567              |cfgloop.c:1567

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2012-04-16
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-04-16 09:32:47 UTC ---
Mine.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-04-16 10:58:06 UTC ---
Looks like a pre-existing bug to me - when we unswitch loops computing
irreducible regions before / after that can lead to inconsistent results.

Loop preserving leads to different loop structure at the beginning of
RTL unswitching compared to the 4.7 branch.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mpolacek at gcc dot gnu.org
         AssignedTo|rguenth at gcc dot gnu.org  |mpolacek at gcc dot gnu.org

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> 2012-11-28 09:30:09 UTC ---
Reassigning.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #4 from Marek Polacek <mpolacek at gcc dot gnu.org> 2012-12-20 13:37:47 UTC ---
The issue here is that when unswitching, we create this new bb:
;; basic block 19, loop depth 0, count 0, freq 14, maybe hot
;;  prev block 20, next block 1, flags: (NEW, RTL, MODIFIED)
;;  pred:       18 [2.2%] 
;;  succ:       8 [100.0%]  (FALLTHRU)

its loop depth should be 1, not 0.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 10:30:54 UTC ---
It no longer reproduces for me ... we no longer unswitch.  But the issue
is certainly latent.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-checking
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 10:33:31 UTC ---
Fixed for this testcase.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 10:32:51 UTC ---
Author: rguenth
Date: Fri Dec 21 10:32:43 2012
New Revision: 194659

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194659
Log:
2012-12-21  Richard Biener  <rguenther@suse.de>

    PR tree-optimization/52996
    * gcc.dg/torture/pr52996.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr52996.c
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
                 CC|                            |jakub at gcc dot gnu.org
         Resolution|FIXED                       |

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-21 12:31:47 UTC ---
The testcase fails on i686-linux:
/usr/src/gcc/obj479/gcc/xgcc -B/usr/src/gcc/obj479/gcc/
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c -fno-diagnostics-show-caret
-O2 -funswitch-loops -S -o pr52996.s
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c: In function 'fn1':
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c:32:1: error: edge from 18
to 19 should be marked irreducible
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c:32:1: error: basic block 19
should be marked irreducible
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c:32:1: error: edge from 19
to 8 should be marked irreducible
/usr/src/gcc/gcc/testsuite/gcc.dg/torture/pr52996.c:32:1: internal compiler
error: in verify_loop_structure, at cfgloop.c:1582
0x82eec96 verify_loop_structure()
        ../../gcc/cfgloop.c:1582
0x856db5d unswitch_loops()
        ../../gcc/loop-unswitch.c:151
0x8558640 rtl_unswitch
        ../../gcc/loop-init.c:336
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.

FAIL: gcc.dg/torture/pr52996.c  -O2  (internal compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O2  (test for excess errors)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer  (internal compiler
error)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer  (test for excess
errors)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer -funroll-loops 
(internal compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer -funroll-loops  (test
for excess errors)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer -funroll-all-loops
-finline-functions  (internal compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O3 -fomit-frame-pointer -funroll-all-loops
-finline-functions  (test for excess errors)
FAIL: gcc.dg/torture/pr52996.c  -O3 -g  (internal compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O3 -g  (test for excess errors)
FAIL: gcc.dg/torture/pr52996.c  -O2 -flto -flto-partition=none  (internal
compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O2 -flto -flto-partition=none  (test for
excess errors)
FAIL: gcc.dg/torture/pr52996.c  -O2 -flto  (internal compiler error)
FAIL: gcc.dg/torture/pr52996.c  -O2 -flto  (test for excess errors)

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 12:47:54 UTC ---
Ick, I had

Index: gcc/cprop.c
===================================================================
--- gcc/cprop.c (revision 194658)
+++ gcc/cprop.c (working copy)
@@ -1554,8 +1554,7 @@ bypass_block (basic_block bb, rtx setcc,
       /* The irreducible loops created by redirecting of edges entering the
         loop from outside would decrease effectiveness of some of the
         following optimizations, so prevent this.  */
-      if (may_be_loop_header
-         && !(e->flags & EDGE_DFS_BACK))
+      if (may_be_loop_header)
        {
          ei_next (&ei);
          continue;

in my tree ...

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |ASSIGNED
         AssignedTo|mpolacek at gcc dot gnu.org |rguenth at gcc dot gnu.org

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 12:55:59 UTC ---
let me take it - bypassing of a loop header is never a good idea unless you
want to jump through hoops as the gimple threading code does.

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 14:34:08 UTC ---
Author: rguenth
Date: Fri Dec 21 14:33:59 2012
New Revision: 194666

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194666
Log:
2012-12-21  Richard Biener  <rguenther@suse.de>

    PR rtl-optimization/52996
    * cprop.c (bypass_block): When loops are to be preserved
    do not bypass loop headers.  Revert earlier kludge to remove
    loops when doing that.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cprop.c

[Bug middle-end/52996] [4.8 Regression] ice in verify_loop_structure, at cfgloop.c:1567

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52996

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

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

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-21 14:35:07 UTC ---
Fixed.