[Bug middle-end/50335] New: ICE in psct_dynamic_dim, at graphite-poly.h:659

[Bug middle-end/50335] New: ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

             Bug #: 50335
           Summary: ICE in psct_dynamic_dim, at graphite-poly.h:659
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: gmaxwell@gmail.com


Created attachment 25231
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25231
preprocessed case that reproduces it

Compiling libopus
http://git.xiph.org/?p=opus.git;a=snapshot;h=f329fa6f0f1e08506a210608bc6de34b5949ec6b;sf=tgz

libcelt/pitch.c: In function 'pitch_search':
libcelt/pitch.c:152:6: internal compiler error: in psct_dynamic_dim, at
graphite-poly.h:659

$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/home/gmaxwell/gcc/libexec/gcc/x86_64-unknown-linux-gnu/4.7.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ./configure --enable-lto --prefix=/home/gmaxwell/gcc/
--enable-cloog-backend=isl --disable-java --disable-libgcj --enable-lipo
--enable-graphite --enable-languages=c --disable-bootstrap
--enable-stage1-languages=c --disable-libstdcxx-v3
Thread model: posix
gcc version 4.7.0 20110908 (experimental) (GCC) 

CFLAGS='-Ofast -floop-flatten' ; make

(doesn't fail with -O3 -floop-flatten)

Preprocessed file attached.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #1 from Gregory Maxwell <gmaxwell at gmail dot com> 2011-09-17 00:31:43 UTC ---
Still failing in r178925.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #2 from Gregory Maxwell <gmaxwell at gmail dot com> 2011-10-07 14:52:37 UTC ---
Still failing in r179659.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #3 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-07 14:57:11 UTC ---
No miracles, eh?

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #4 from Gregory Maxwell <gmaxwell at gmail dot com> 2011-10-07 15:29:42 UTC ---
Ha, indeed. Well, I don't want it to be forgotten just because its become old
and potentially inapplicable.

I test GCC development periodically on all the software I work on— most of the
time anything I find is fixed quickly, it's not something urgent for me. (I
once reported a bug in a proprietary compiler: reporting bugs in GCC is my
penance ;) )

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #5 from Gregory Maxwell <gmaxwell at gmail dot com> 2011-11-14 06:19:44 UTC ---
Still failing with r181345

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #6 from Gregory Maxwell <gmaxwell at gmail dot com> 2011-12-06 04:51:47 UTC ---
Still failing in r182038

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[mkuvyrkov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

Maxim Kuvyrkov <mkuvyrkov at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mkuvyrkov at gcc dot
                   |                            |gnu.org, spop at gcc dot
                   |                            |gnu.org

--- Comment #7 from Maxim Kuvyrkov <mkuvyrkov at gcc dot gnu.org> 2011-12-06 08:50:37 UTC ---
Sebastian,

Putting this PR on your horizon.  This is an easily reproducible ICE in
graphite-poly.h.

Let us know if you are swamped and won't be able to look into this.

Thank you.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[sebpop at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #8 from sebpop at gmail dot com <sebpop at gmail dot com> 2011-12-06 18:40:43 UTC ---
Hi Maxim,

Thanks for pointing me to this problem.

I would recommend that you disable the code in loop flattening by
early returning false in flatten_all_loops.  I'm pre-approving you to
commit such a patch (as I am not allowed to commit anything to gcc
due to my new employer's legal policies, although emailing or giving
instructions via a phone call to somebody of what to do seems to
be fine for now ;-)

Vladimir Kargov will take over where I left the patches to make graphite
work with ISL and once these are committed to gcc, we will be able to
implement loop flattening with ISL, by calling isl_map_flatten_range.

Sebastian

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[maxim_kuvyrkov at mentor dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #9 from maxim_kuvyrkov at mentor dot com 2011-12-08 00:06:40 UTC ---
On 7/12/2011, at 7:40 AM, sebpop at gmail dot com wrote:

> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335
> 
> --- Comment #8 from sebpop at gmail dot com <sebpop at gmail dot com> 2011-12-06 18:40:43 UTC ---
> Hi Maxim,
> 
> Thanks for pointing me to this problem.
> 
> I would recommend that you disable the code in loop flattening by
> early returning false in flatten_all_loops.

Do I understand correctly that you are suggesting effectively disabling loop
flattening completely?

Thank you,

--
Maxim Kuvyrkov
Mentor Graphics

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[sebpop at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #10 from sebpop at gmail dot com <sebpop at gmail dot com> 2011-12-08 04:08:14 UTC ---
On Wed, Dec 7, 2011 at 18:06, maxim_kuvyrkov at mentor dot com
<gcc-bugzilla@gcc.gnu.org> wrote:
> Do I understand correctly that you are suggesting effectively disabling loop
> flattening completely?

Yes, that would fix this problem and other potential bugs
related to this implementation of flattening.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #11 from Gregory Maxwell <gmaxwell at gmail dot com> 2012-02-14 17:15:23 UTC ---
Still crashing in, r184217.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #12 from Gregory Maxwell <gmaxwell at gmail dot com> 2012-02-22 09:19:54 UTC ---
_Not_ failing in r184458.  Perhaps fixed by r184265?

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[gmaxwell at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

--- Comment #13 from Gregory Maxwell <gmaxwell at gmail dot com> 2012-02-22 09:39:51 UTC ---
Okay, it doesn't ICE now— but now I have cases where it apparently runs forever
and cases where it produces incorrect code with -floop-flatten.

[Bug middle-end/50335] ICE in psct_dynamic_dim, at graphite-poly.h:659

[mkuvyrkov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50335

Maxim Kuvyrkov <mkuvyrkov at gcc dot gnu.org> changed:

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

--- Comment #14 from Maxim Kuvyrkov <mkuvyrkov at gcc dot gnu.org> 2012-03-02 20:36:12 UTC ---
Fixed by removing loop flattening optimization.  The -floop-flatten flag no
does nothing.