[vtkusers] efficiency

[Lisa Sobierajski Avila]

Hello Jim,

I have been tuning the texture mapper a bit, and I still have some more 
tweaking to go. What is likely to improve is the rendering speed for 
relatively small volumes that don't use shading or gradient magnitude 
opacity transfer functions (that is because each of those require an extra 
lookup). The rendering performance of large volumes is not likely to 
improve in the near future. I don't think the bottleneck is OO structures 
or timing the rendering process - I believe it is more that we have not 
tuned it to a particular graphics card, and we have not limited the 
functionality. I am sure that better performance can be obtained if you 
know in advance that you won't do shading / gradient opacity, you know that 
your volume is a power of two, you are willing to store 3 copies of your 
volume in RGBA, you know the "right" settings for your card to make it work 
well, etc.

If anyone out there has some generic suggestions on how to improve the 2D 
texture mapper, I'd be happy to hear about them.


Lisa



At 02:50 AM 6/29/00, Jim Morash wrote:
>Hello,
>
>I'm using VTK to do volume rendering of a CT dataset. The priority in
>this case is to render the dataset (with variable opacity, currently
>256^3 but ideally 512^3) as fast as possible (ideally 10+ fps). I'm
>currently using vtkVolumeTextureMapper2D, which works decently well but
>is just not fast enough (one frame takes about 0.78s). Does anyone have
>any comments about how much computational overhead vtk incurs by having
>such a nice neat OO structure and all kinds of extras available (such as
>timing the rendering process, etc.)?  Would it be worth the (large
>amount of) effort to try and implement my own volume renderer, or is vtk
>really pretty lean already?
>Any advice would be very much appreciated... since we're trying to
>produce something that would be cheap for institutions to acquire,
>specialized hardware like a VolumePro is on the "we'd rather not" end of
>the spectrum.
>--Jim Morash
>
>jmorash at mit.edu                     http://web.mit.edu/jmorash/www
>------------------------------------------------------------------
>Loyalty to petrified opinions never yet broke a chain or freed a
>human soul in this world--and never will. --Mark Twain
>
>_______________________________________________
>This is the private VTK discussion list.
>Please keep messages on-topic. Check the FAQ at: 
><http://public.kitware.com/cgi-bin/vtkfaq>
>vtkusers mailing list
>vtkusers at public.kitware.com
>http://public.kitware.com/mailman/listinfo/vtkusers