[Paraview-developers] Using vtkImageMarchingCubes as multithread process gives wrong output

[Utkarsh Ayachit]

The problem seems to be that the vtkImageMarchingCubes filter is not
designed to be used in cases where data is also partitioned among
ranks. It's RequestData(..) implementation directly uses WHOLE_EXTENT
to determine chunking for threaded executing, it should have been
using UPDATE_EXTENT.

Utkarsh

On Thu, Oct 23, 2014 at 8:45 AM, Frank TT <frank.tt at einseinself.org> wrote:
> On 2014-10-22 20:58, Utkarsh Ayachit wrote:
>>
>> Frank,
>>
>> Did you forget to attach the attachments.
>
>
> Oh, yes. Sorry.
>
>
>
>
>
>>
>> Utkarsh
>>
>> On Wed, Oct 22, 2014 at 6:40 AM, Frank tt <frank.tt at einseinself.org>
>> wrote:
>>>
>>> Hi,
>>>
>>> I wrote a PV plugin based on vtkImageMarchingCubes.h. At my local
>>> machine it's working fine. But at the server, when running with mpiexec
>>> I get some wrong output of the createt surface. (See attachments)
>>>
>>> Now I'm not sure if this depends on the vtkImageMarchingCubes.h or at
>>> mpiexec.
>>>
>>> I'm using at server:
>>> * cuda-5.5
>>> * gcc-4.7.1
>>> * mesa-7.11.2 (selfbuild, like the superbuild)
>>> * openmpi-1.8.1 (selfbuild)
>>> * paraview-4.1.0 (selfbuild)
>>>
>>> Has anybody any idea what I did wrong or what I can do?
>>>
>>> I have the option to use the servers mpi installation
>>> (parastation/gcc-mt-5.0.29). This is what I'm actually doing (I'm
>>> waiting for the next proc slot a the server).
>>>
>>> Greetings
>>>
>>> Frank
>>>
>>> _______________________________________________
>>> Paraview-developers mailing list
>>> Paraview-developers at paraview.org
>>> http://public.kitware.com/mailman/listinfo/paraview-developers