Bastil

I cannot comment on the surface streamlines directly, as I'm not familiar with the class that is used...but if it internally makes use of the vtkStreamTracer filter code, then I can make the following observation.

The vtkStreamTracer class uses vtkDataSet FindCell to locate the start point of a given streamline. It then builds links (vtkUnstructuredGrid only) between cells so that as a streamline passes out of one cell and onto the next, the neighbour can be located quickly. The findCell is very slow as for an unstructuredgrid, it is a brute force search of all cells. The build links is also very slow and if your grid is a single 20million cell dataset, it will take ages.

About a year ago, I posted a request to this list for volunteers, to test some improvements that I'd made to the StreamTracer code to
a) make use of a cell locator instead of FindCell/BuildLinks to accelerate the streamline searches.
b) Clean up some caching to prevent certain operations taking place on every iteration of the inner integration loop within the streamline calculation.

I got no volunteers, so I never committed my code. Since then I had problems with maintenance of my tree and reverted all my changes (I was only interested in my TemporalStreamTracer code anyway), but a few weeks ago, I decided to re-tweak the streamline filter to see if my new CachingInterpolatedVelocityField would work with it. I found that my own streamline generation was greatly improved (20million cells approx in my case also).

The changes are in the pv-meshless tree (use google), but I will not offer further assistance as I'm very busy this week. So if you want to try it out, you are welcome - and if you can post results to this list, then so much the better - and it may prompt me into a second round of fixes. (the changes I made a few weeks ago were not thorough and I need to spend a little more time on it before it's ready to be committed).

JB
Hi group,

I tried creating some surface streamlines as described in the wiki.
Process seemed to work but was extremly slow for me so I could not wait it.
I have large CFD models imported as ensight (~ 20 Million cells).
Paraview run for about five hours without an result for surface
streamlines from 50 seed points. I don't really know if this is normal?

Regards BastiL
_______________________________________________
Powered by www.kitware.com

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview
  


-- 
John Biddiscombe,                            email:biddisco @ cscs.ch
http://www.cscs.ch/
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Cantonale, 6928 Manno, Switzerland      | Fax:  +41 (91) 610.82.82
_______________________________________________
Powered by www.kitware.com

Visit other Kitware open-source projects at 
http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: 
http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview

Reply via email to