[vtkusers] MPI pvti num_pieces >> num_procs

[Kevin H. Hobbs]

I want to use our cluster and stream a huge image through the
vtkImageAnisotropicDiffusion3D filter and write it to a pvti file.  I've
attached the code that should do that. The trouble is every node uses a
huge amount of RAM and eventually dies because it used up all the RAM
and swap. The number of pieces is set large enough that each piece
should be small. When I use similar code that runs on only one computer
and uses the serial XML image writer the memory usage is small and the
pieces come out every few seconds.

Something about the 'pvti' writer in VTK from CVS is loading WAY too
much of the image.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: MPIImageAnisotropicDiffusion3D.cxx
Type: text/x-c++src
Size: 2533 bytes
Desc: not available
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20060922/0b61e24f/attachment.cxx>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20060922/0b61e24f/attachment.pgp>