[vtkusers] isosurfacing large volumes in parallel

[Dominik Szczerba]

On Wednesday 23 January 2008 12.14:13 Kevin H. Hobbs wrote:
> 
> On Tue, 2008-01-22 at 09:23 +0100, Dominik Szczerba wrote:
> > Thank you for a hint. Indeed, there is an explicit example for 
isosurfacing.
> > Any idea if the same will work for vtkClipVolume? The whole volume does 
not 
> > fit into one PC. In particular, I am worried about edge effects, i.e., the 
> > generated mesh must be compatible on domain cuts.
> > thanks a lot for any hints,
> > Dominik
> 
> You should be able to do this. take a look at ghost levels for the
> unstructured grid writers. Note that clip produces an unstructured grid,
> and cut produces a polydata.

Currently I did a (dis)proof of concept by simply generating a scalar field 
consistently on one 3D ImageData (signed distance function), _THEN_ cutting 
it along one axis to produce domains (so the field is consistent and no edge 
effects are produced), then clipping each of them separately. I find that 
neither the nodes nor the elements are compatible on the domains boundaries. 
I triple made sure I am not missing one layer of pixels by mistake. Will the 
distributed version work in such a case to ensure compatibility? In the code 
for vtkClipVolume I did not find any special treatment for the distributed 
case, like communication or ghost cells.

thanks a lot,
Dominik

-- 
Dominik Szczerba, Ph.D.
Computer Vision Lab CH-8092 Zurich
http://www.vision.ee.ethz.ch/~domi
-------------- 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/20080123/51b564bc/attachment.pgp>