[petsc-users] Optimized run crashes on one machine but not another

[Garnet Vaz]

Hi Matt,

I figured out what was causing the problems. I never had
an optimized build built on my laptop before. So I did not
run the reconfigure script but had to build from scratch.

I recently changed my build script to keep only what was
required for this project. By mistake, the --download-chaco
was taken out. Hence, the code was running on all debug
builds + optimized build on the desktop and crashed on
my laptop.

Once I rebuilt 3.4.2 it crashed for all builds since chaco
was not available on any of them now. The IS of
the partition returned here was corrupt.
plex.c line 2503
    if (1) {
#if defined(PETSC_HAVE_CHACO)
      ierr = DMPlexPartition_Chaco(dm, numVertices, start, adjacency,
partSection, partition);CHKERRQ(ierr);
#endif
    } else {
#if defined(PETSC_HAVE_PARMETIS)
      ierr = DMPlexPartition_ParMetis(dm, numVertices, start, adjacency,
partSection, partition);CHKERRQ(ierr);
#endif
    }
I do have parmetis installed but I do not know why it did not
work.

Since you were able to run the code it helped a lot. The
code runs on all builds/machines now without any problems
once chaco was downloaded.

Thanks a lot for the help.

Regards,
Garnet




On Wed, Aug 28, 2013 at 7:22 PM, Matthew Knepley <knepley at gmail.com> wrote:

> On Wed, Aug 28, 2013 at 4:53 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>
>> Garnet Vaz <garnet.vaz at gmail.com> writes:
>>
>> > Hi Matt,
>> >
>> > Within gdb how can I view an IS? I tried 'call ISView(*partition,0)'
>> > following the VecView() syntax but it causes a segmentation fault
>> > inside gdb.
>>
>> That'll work if you're passing an IS and it's not already corrupt.
>>
>
> Just to recap, I have run in both debug and optimized with complex, and
> through valgrind,
> and I get no problems on my machine (using downloaded MPICH).
>
> I think we will need you to narrow it down with the debugger there.
>
>   Thanks,
>
>      Matt
>
>
> --
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> -- Norbert Wiener
>



-- 
Regards,
Garnet
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130828/7844e1cd/attachment.html>