[vtkusers] MPI and vtkDiscreteMarchingCubes, vtkWindowedSincPolyDataFilter

[Grothausmann, Roman Dr.]

Dear mailing list members,


I've created a small program to combine vtkDiscreteMarchingCubes, 
vtkWindowedSincPolyDataFilter and a check whether the result is closed and 
whether it has non-manifold edges:
https://github.com/romangrothausmann/VTK-CLIs/blob/master/discrete_marching-cubes.cxx
This works find and as expected but is very slow for our current 30GB input 
datasets. So I tried to parallelize it with MPI. However both tries do not work 
as expected. One is based on a VTK-workshop example 
ParallelCompositeTime/Parallel/parallel1_answer.cxx:
https://github.com/romangrothausmann/VTK-CLIs/blob/dmc_MPI/discrete_marching-cubes_MPI.cxx
which does not crash but apparently does not split the data even though 
SetUpdateExtent is used (written files all contain the whole data).
The other based on VTK's ParallelIso.cxx example:
https://github.com/romangrothausmann/VTK-CLIs/blob/dmc_MPI_t02/discrete_marching-cubes_MPI.cxx
Seems to do the same but additionally segfauls on some sub-processes.
What am I doing wrong?
Where can I check whether a VTK-filter can stream, i.e. does 
vtkWindowedSincPolyDataFilter only need ghost levels or the whole data?

Sadly, neither adding UpdateInformation() nor changing the reader (from 
vtkMetaImageReader (marching-cubes_MPI_01.cxx) to vtkPNrrdReader 
(marching-cubes_MPI_02.cxx)) solved the problem. I also reduced my exemplary try 
to just use vtkContourFilter, as in IO/MPIImage/Testing/Cxx/ParallelIso.cxx.
Does vtkMetaImageReader not support MPI/piecewise extents?
Why is there no call to UpdateInformation() nor SetUpdateExtent() in 
ParallelIso.cxx (I find it a quite complex example)?

Thanks for any help or hints.
Roman
-- 
Dr. Roman Grothausmann

Tomographie und Digitale Bildverarbeitung
Tomography and Digital Image Analysis

Institut für Funktionelle und Angewandte Anatomie, OE 4120
Medizinische Hochschule Hannover
Carl-Neuberg-Str. 1
D-30625 Hannover

Tel. +49 511 532-2900