[vtkusers] How to visualize thousands of actors with good performance

[Karthik Krishnan]

2010/9/29 Affolter Martin <martin.affolter at ntb.ch>

>  The example I used for comparison has about 8000 actors, but we also have
> bigger models with several 10'000 actors. I honestly doubt, that our problem
> is OpenGL related. We've been using VTK 4.2 and VTK 4.4 for some years now
> and have never had performance issues so far. Those only emerged after we
> moved to VTK 5.6.


Interesting.

There are two *major* changes between VTK4.4 and now.
- The painter infrastructure. (introduced in VTK-5-2)
- New pipeline. (Introduced in VTK-5-0)

An interesting exercise would be to update to VTK-5-0 (before the painters
but after the new pipeline) and check the performance.

--
karthik


> Both versions use OpenGL, but only with 5.6 the interaction performance is
> poor.
>
> So far I've tried static mappers and/or immediatemoderendering with little
> success. Right now I'm looking for a way to file an example that I can
> provide to those willing to look at it. OpenSceneGraph sounds interesting.
>
> Regards
> Martin
>
> Von: "Björn Zehner" [mailto:bjoern.zehner at ufz.de]
> Gesendet: Dienstag, 28. September 2010 18:00
> An: Affolter Martin
> Cc: vtkusers at vtk.org
> Betreff: Re: [vtkusers] How to visualize thousands of actors with good
> performance
>
> Hello Martin,
>
> how many thousands of actors do you have? I am asking because rendering
> performance is not only limited by the number of vertices and triangles you
> have but by the number of objects too. This has more to do with OpenGL than
> with VTK. I think this is explained in more detail in:
> "Cebenoyan, C. (2004): Graphics Pipeline Performance. In: GPU Gems, Edited
> by Randima Fernando, Addison Wesley, Boston"
> I believe that NVidia has made this available online.
>
> If this is your problem, than you need some datastructure that allows to
> efficiently cull away the objects that are not visible (e.g. a quadtree).
> Are all objects (vtkActors) visible at the same time? If yes, you might need
> to make some assumptions, such as that the users can only pick objects that
> are near, and combine these culling techniques with Level Of Detail
> techniques. I am not sure if such an implementation can be done with VTK,
> may be you need a scenegraph that is targeted at rendering, such as OpenSG
> or OpenSceneGraph.
>
> Kind regards,
> Björn Zehner
>
>
> -------------------------------------------------------------------------
> Dr. Bjoern Zehner
> UFZ Centre for Environmental Research Leipzig-Halle
> Permoserstrasse 15
> 04318 Leipzig
> Germany
> http://www.ufz.de/index.php?en=5673
> Tel: ++49 (341) 235 1979
> Fax: ++49 (341) 235 1939
> _______________________________________________
> 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 VTK FAQ at:
> http://www.vtk.org/Wiki/VTK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtkusers
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20100929/86a4f248/attachment.htm>