[Bug c++/53672] New: gcc/branches/cilkplus does not like C++ spawn (when optimized)

[Bug c++/53672] New: gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

             Bug #: 53672
           Summary: gcc/branches/cilkplus does not like C++ spawn (when
                    optimized)
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: jjhforrest@gmail.com


Created attachment 27619
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27619
.ii file created with -save-temp

In real code I get

In function '<built-in>':
cc1plus: internal compiler error: Segmentation fault

With tiny example (segfault.cpp -> segfault.ii) I get

/j/mycilk3/bin/g++ -O2 segfault.ii
segfault.cpp: In function ‘<built-in>’:
segfault.cpp:33:1: internal compiler error: in remap_gimple_op_r, at
tree-inline.c:849
 }
 ^
works at -O1

Output from  /j/mycilk3/bin/g++ -v
Using built-in specs.
COLLECT_GCC=/j/mycilk3/bin/g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-unknown-linux-gnu/4.8.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../cilk2/configure --prefix=/j/mycilk3 --libexecdir=/usr/lib
--enable-__cxa_atexit --enable-clocale=gnu --enable-languages=c,c++
--disable-multilib --disable-bootstrap --with-system-zlib CFLAGS='-fPIC
-march=x86-64' LD=/usr/bin/ld
Thread model: posix
gcc version 4.8.0 20120522 (experimental) (GCC)

segfault.ii attached

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[bviyer at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

Balaji V. Iyer <bviyer at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bviyer at gmail dot com

--- Comment #1 from Balaji V. Iyer <bviyer at gmail dot com> 2012-06-14 16:56:19 UTC ---
Hello John,
   This problem seem to be fixed as of this commit:

git:0ac59c91905a106865589114d4e55f0c7f256874
svn: revision:188147

I tried your code and mine passes fine.

Thanks,

Balaji V. Iyer.

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #2 from John Forrest <jjhforrest at gmail dot com> 2012-06-14 19:22:30 UTC ---
Balaji,

Updated my copy.

My simple version - now works - but real case still doesn't.

Try and give you a .ii file tomorrow.


John


On 14/06/12 17:56, bviyer at gmail dot com wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672
>
> Balaji V. Iyer<bviyer at gmail dot com>  changed:
>
>             What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                   CC|                            |bviyer at gmail dot com
>
> --- Comment #1 from Balaji V. Iyer<bviyer at gmail dot com>  2012-06-14 16:56:19 UTC ---
> Hello John,
>     This problem seem to be fixed as of this commit:
>
> git:0ac59c91905a106865589114d4e55f0c7f256874
> svn: revision:188147
>
> I tried your code and mine passes fine.
>
> Thanks,
>
> Balaji V. Iyer.
>

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #3 from John Forrest <jjhforrest at gmail dot com> 2012-06-15 06:57:43 UTC ---
Balaji,

Attached - AbcMatrix.ii gives segfault using "compile"

There are more than a dozen places in my code where code such as -

>    for (int i=0;i<numberBlocks-1;i++)
>      which[i]=cilk_spawn pivotColumnDantzig(i,useRowCopy,updates,spare,best[i]);
>    which[numberBlocks-1]=pivotColumnDantzig(numberBlocks-1,useRowCopy,updates,
> 						       spare,best[numberBlocks-1]);
>    cilk_sync;
>

  inside a class such as AbcMatrix segfaults on compilation.  If I 
create a non class static function doWork(AbcMatrix * matrix, other 
stuff) and in that function do

which[i]=cilk_spawn(matrix,other stuff)

it works.  There is not much of an overhead but is ugly and time consuming.

Hope you can reproduce fault.

Regards,

John

On 14/06/12 17:56, bviyer at gmail dot com wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672
>
> Balaji V. Iyer<bviyer at gmail dot com>  changed:
>
>             What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                   CC|                            |bviyer at gmail dot com
>
> --- Comment #1 from Balaji V. Iyer<bviyer at gmail dot com>  2012-06-14 16:56:19 UTC ---
> Hello John,
>     This problem seem to be fixed as of this commit:
>
> git:0ac59c91905a106865589114d4e55f0c7f256874
> svn: revision:188147
>
> I tried your code and mine passes fine.
>
> Thanks,
>
> Balaji V. Iyer.
>

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[bviyer at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #4 from bviyer at gcc dot gnu.org 2012-06-15 19:44:04 UTC ---
Author: bviyer
Date: Fri Jun 15 19:43:57 2012
New Revision: 188680

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188680
Log:
This patch fixes bug PR 53672.

2012-06-15  Balaji V. Iyer  <balaji.v.iyer@intel.com>  

    * dwarf2out.c (dwarf2out_function_decl): Added a check for spawn helper.



Modified:
    branches/cilkplus/gcc/ChangeLog.cilk
    branches/cilkplus/gcc/dwarf2out.c

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[bviyer at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #5 from Balaji V. Iyer <bviyer at gmail dot com> 2012-06-15 19:46:13 UTC ---
Hello John,
    The revision mentioned above should fix this problem.

Thanks,

Balaji V. Iyer.

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #6 from John Forrest <jjhforrest at gmail dot com> 2012-06-16 08:13:02 UTC ---
Balaji,

Thanks - seems to fix bug - now I can simplify my coding in various places.

John

On 15/06/12 20:46, bviyer at gmail dot com wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672
>
> --- Comment #5 from Balaji V. Iyer<bviyer at gmail dot com>  2012-06-15 19:46:13 UTC ---
> Hello John,
>      The revision mentioned above should fix this problem.
>
> Thanks,
>
> Balaji V. Iyer.
>

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #7 from John Forrest <jjhforrest at gmail dot com> 2012-06-16 16:08:09 UTC ---
Sorry to say that when I was backing out some of my previous 
work-arounds I got a segfault again.

John

On 15/06/12 20:46, bviyer at gmail dot com wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672
>
> --- Comment #5 from Balaji V. Iyer<bviyer at gmail dot com>  2012-06-15 19:46:13 UTC ---
> Hello John,
>      The revision mentioned above should fix this problem.
>
> Thanks,
>
> Balaji V. Iyer.
>

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[bviyer at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #8 from bviyer at gcc dot gnu.org 2012-06-18 02:08:03 UTC ---
Author: bviyer
Date: Mon Jun 18 02:07:54 2012
New Revision: 188722

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188722
Log:
This patch should fix the 2nd case of PR 53672.
2012-06-17  Balaji V. Iyer  <balaji.v.iyer@intel.com>

    PR 53672
    * dwarf2out.c (dwarf2out_abstract_function): Added a check for the
    spawn helper.
    (dwarf2out_function_decl): Added a check if CILK_FN_P is set when there
    is no function struct.
    * cilk.c (expand_builtin_cilk_metadata): Set the CILK_FN_P field to 1.
    * cilk-spawn.c (cg-hacks): Likewise.
    (gimplify_cilk_for_2): Likewise.
    (gimplify_cilk_for): Likewise.
    * tree.h (CILK_FN_P): Added new #define to access a base field of tree.


Modified:
    branches/cilkplus/gcc/ChangeLog.cilk
    branches/cilkplus/gcc/cilk-spawn.c
    branches/cilkplus/gcc/cilk.c
    branches/cilkplus/gcc/cp/ChangeLog.cilk
    branches/cilkplus/gcc/cp/cilk.c
    branches/cilkplus/gcc/dwarf2out.c
    branches/cilkplus/gcc/tree.h

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[jjhforrest at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

--- Comment #9 from John Forrest <jjhforrest at gmail dot com> 2012-06-18 09:59:11 UTC ---
Balaji,

That works and on a few other cases I have altered back.

Hopefully that is that.

Thanks

John
On 18/06/12 03:08, bviyer at gcc dot gnu.org wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672
>
> --- Comment #8 from bviyer at gcc dot gnu.org 2012-06-18 02:08:03 UTC ---
> Author: bviyer
> Date: Mon Jun 18 02:07:54 2012
> New Revision: 188722
>
> URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188722
> Log:
> This patch should fix the 2nd case of PR 53672.
> 2012-06-17  Balaji V. Iyer<balaji.v.iyer@intel.com>
>
>      PR 53672
>      * dwarf2out.c (dwarf2out_abstract_function): Added a check for the
>      spawn helper.
>      (dwarf2out_function_decl): Added a check if CILK_FN_P is set when there
>      is no function struct.
>      * cilk.c (expand_builtin_cilk_metadata): Set the CILK_FN_P field to 1.
>      * cilk-spawn.c (cg-hacks): Likewise.
>      (gimplify_cilk_for_2): Likewise.
>      (gimplify_cilk_for): Likewise.
>      * tree.h (CILK_FN_P): Added new #define to access a base field of tree.
>
>
> Modified:
>      branches/cilkplus/gcc/ChangeLog.cilk
>      branches/cilkplus/gcc/cilk-spawn.c
>      branches/cilkplus/gcc/cilk.c
>      branches/cilkplus/gcc/cp/ChangeLog.cilk
>      branches/cilkplus/gcc/cp/cilk.c
>      branches/cilkplus/gcc/dwarf2out.c
>      branches/cilkplus/gcc/tree.h
>

[Bug c++/53672] gcc/branches/cilkplus does not like C++ spawn (when optimized)

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53672

Paolo Carlini <paolo.carlini at oracle dot com> changed:

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

--- Comment #10 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-10-10 19:34:17 UTC ---
Closing.