[Bug regression/56738] New: ICE in c-c++-common/torture/vshuf-v4di.c

[Bug regression/56738] New: ICE in c-c++-common/torture/vshuf-v4di.c

[ktkachov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

             Bug #: 56738
           Summary: ICE in c-c++-common/torture/vshuf-v4di.c
    Classification: Unclassified
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: regression
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ktkachov@gcc.gnu.org
                CC: stevenb.gcc@gmail.com
            Target: arm-none-eabi


Created attachment 29729
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29729
Preprocessed file

Testcase ICEs on arm-none-eabi when compiled with -O1 (ICEs also with -O2 and
-O3)

In file included from vshuf-v4di.c:19:0:
vshuf-main.inc: In function 'main':
vshuf-main.inc:26:1: internal compiler error: in df_insn_delete, at
df-scan.c:1162
 }
 ^
0x634b9c df_insn_delete(rtx_def*)
    $SOURCE/gcc/gcc/df-scan.c:1162
0x68b98a remove_insn(rtx_def*)
    $SOURCE/gcc/gcc/emit-rtl.c:4036
0x5f83be delete_insn(rtx_def*)
    $SOURCE/gcc/gcc/cfgrtl.c:167
0xd335b6 resolve_simple_move
    $SOURCE/gcc/gcc/lower-subreg.c:1072
0xd338b3 resolve_simple_move
    $SOURCE/gcc/gcc/lower-subreg.c:923
0xd34a13 decompose_multiword_subregs
    $SOURCE/gcc/gcc/lower-subreg.c:1563
0xd3528d rest_of_handle_lower_subreg2
    $SOURCE/gcc/gcc/lower-subreg.c:1682
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.


Bisection shows it started with r196977

Cross compiler configured with:
Target: arm-none-eabi
Configured with: $SOURCE/gcc/configure --target=arm-none-eabi
--prefix=$SOURCE/build/install --with-gmp=$SOURCE/build/host-tools
--with-mpfr=$SOURCE/build/host-tools --with-mpc=$SOURCE/build/host-tools
--with-pkgversion=unknown --disable-shared --disable-nls --disable-threads
--disable-tls --enable-checking=yes --enable-languages=c --with-newlib
--with-fpu=neon --with-float=hard --with-arch=armv7-a
Thread model: single
gcc version 4.9.0 20130322 (experimental) (unknown)

[Bug regression/56738] ICE in c-c++-common/torture/vshuf-v4di.c

[ktkachov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

ktkachov at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.9.0

[Bug regression/56738] ICE in c-c++-common/torture/vshuf-v4di.c

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

--- Comment #1 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-03-26 11:57:58 UTC ---
I see a similar error on powerpc-*-* (see
http://gcc.gnu.org/ml/gcc-testresults/2013-03/msg02572.html ) for
c-c++-common/torture/vshuf-v2di.c and g++.dg/torture/vshuf-v2di.C.

Should I fill a new PR or extend this one to vshuf-v2di.c?

[Bug regression/56738] ICE in c-c++-common/torture/vshuf-v4di.c

[ktkachov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

--- Comment #2 from ktkachov at gcc dot gnu.org 2013-03-26 12:07:07 UTC ---
(In reply to comment #1)
> I see a similar error on powerpc-*-* (see
> http://gcc.gnu.org/ml/gcc-testresults/2013-03/msg02572.html ) for
> c-c++-common/torture/vshuf-v2di.c and g++.dg/torture/vshuf-v2di.C.
> 
> Should I fill a new PR or extend this one to vshuf-v2di.c?

Are you seeing the ICE at the same location? (df-scan.c).
Also, from the testresults page you linked I don't see an ICE for -O2 or -O3.
Do those run successfully?

If you believe they are the same error I think you can add powerpc
to the targets in this PR.

Regards,
Kyrill

[Bug regression/56738] [4.9 Regression] ICE in c-c++-common/torture/vshuf-v4di.c

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|arm-none-eabi               |arm-none-eabi powerpc-*-*
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-03-26
            Summary|ICE in                      |[4.9 Regression] ICE in
                   |c-c++-common/torture/vshuf- |c-c++-common/torture/vshuf-
                   |v4di.c                      |v4di.c
     Ever Confirmed|0                           |1

--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-03-26 12:43:49 UTC ---
> Are you seeing the ICE at the same location? (df-scan.c).

Yes:

In file included from
/opt/gcc/work/gcc/testsuite/g++.dg/torture/vshuf-v2di.C:18:0:
/opt/gcc/work/gcc/testsuite/g++.dg/torture/vshuf-main.inc: In function 'int
main()':
/opt/gcc/work/gcc/testsuite/g++.dg/torture/vshuf-main.inc:29:1: internal
compiler error: in df_insn_delete, at df-scan.c:1162
 }


> Also, from the testresults page you linked I don't see an ICE for -O2 or -O3.
> Do those run successfully?

Yes.

The tests compile with revision 197004, but not with revision 197010. It is
likely revision 197005.

[Bug regression/56738] [4.9 Regression] ICE in c-c++-common/torture/vshuf-v4di.c

[steven at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

Steven Bosscher <steven at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |steven at gcc dot gnu.org

--- Comment #4 from Steven Bosscher <steven at gcc dot gnu.org> 2013-03-27 23:45:01 UTC ---
Bah...

Index: df-scan.c
===================================================================
--- df-scan.c   (revision 197180)
+++ df-scan.c   (working copy)
@@ -1158,8 +1158,17 @@ df_insn_delete (rtx insn)
      In any case, we expect BB to be non-NULL at least up to register
      allocation, so disallow a non-NULL BB up to there.  Not perfect
      but better than nothing...  */
-
+  /* ??? bb can also be NULL if lower-subreg.c:resolve_simple_mov emits
+     an insn into a sequence and then does delete_insn on it.  Not sure
+     if that makes sense, but for now it means this assert cannot work.
+     See PR56738.
+     Disable for now but revisit before the end of GCC 4.9 stage1.  */
+#if 0
   gcc_checking_assert (bb != NULL || reload_completed);
+#else
+  if (bb == NULL)
+    return;
+#endif

   df_grow_bb_info (df_scan);
   df_grow_reg_info ();

[Bug rtl-optimization/56738] [4.9 Regression] ICE in c-c++-common/torture/vshuf-v4di.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING
          Component|regression                  |rtl-optimization

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Is this fixed now?

[Bug rtl-optimization/56738] [4.9 Regression] ICE in c-c++-common/torture/vshuf-v4di.c

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56738

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

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

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.