[Fwd: Multi-Block Performance]

[David D. Marshall]

I sent this message yesterday and this morning, but never saw it on the
mail group.  Is it
even getting through?  I just unsubscribed and resubscribed, maybe that
will help?

"David D. Marshall" wrote:

> First off, I am using an SGI Indigo2 workstation with a 75MHz processor
> and 128 MB of RAM.  I have multi-block data that I have been trying to
> visualize using VTK (color surfaces, streamlines, contours, etc.) and
> quickly realized that the structured grid classes were only intended for
> 1 block.  I then tried (as suggest on the mailing list) to turn my grid
> into unstructured hexahedrons, but I noticed a significant performance
> degradation.  My computational domain consists of 3 blocks 203x35x2,
> 150x45x2, and 150x45x2 for a total of 41,210 points, and it takes 2 3/4
> minutes to bring up the initial view of any visualization technique
> (i.e. cut surfaces, streamlines,etc.).  Only a small fraction of this
> time is spent reading in the data.  I've tried the vtkLODActor, and
> while that increased the panning and zooming time, it had little effect
> on the startup time.  I am now going to look at filtering out grid
> points to bring down the total number of nodes down, but I didn't want
> to do that because our lab has another visualization tool that (FAST and
> Visual3) that can easily handle this grid size.  So my question is this:
> my raw datasets are going to be as large if not larger than this set and
> how cant I improve the performance?
>
> Also there was talk on the mail group that vtkCompositeStructuredGrid
> class might need to be developed to handle multi-block structured
> grids.  I might be interested in developing such a class (as well as
> something similar for unstructured grids) if I was given some pointers
> on how to start (I have The Visualization Toolkit book and The User's
> Guide as references).
>
> --
> David D. Marshall
> ARTLab System Administrator/GRA
> Georgia Institute of Technology, Atlanta Georgia, 30332
> http://www.ae.gatech.edu/research/artlab/artl/artlab.html
> mailto:gte552m at prism.gatech.edu
> think: Why anonymity is good
> http://www.cato.org/pubs/briefs/bp-054es.html
-------------- next part --------------
An embedded message was scrubbed...
From: "David D. Marshall" <gte552m at prism.gatech.edu>
Subject: Multi-Block Performance
Date: Thu, 04 May 2000 11:10:50 -0400
Size: 2425
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20000505/5b383a15/attachment.mht>