Donna L Gresh wrote:
> 
> I gather that you want the specific functionality that maptoplane
> provides in creating just the right triangles that match the
> cubes or tets of the original 3D field? Because you can certainly
> just define a line (using two points to construct) and then refine
> it to a level that works for the data set, then map the 3D data onto it.
> Or are you mapping 2D data in 3 space and that's the issue?
> 
> _________________________________
> Donna L. Gresh, Ph.D.
> Visual Analysis Group
> IBM T.J. Watson Research Center
> (914) 784-5049
> [EMAIL PROTECTED]
> 

I'm looking for the specific functionality of MapToPlane. Basically, I want to
shoot a ray through a dataset (in this case, a 4M Tet CFD simulation). If I
refine to a level fine enough to put at least one sample in a tet at the boundry
layer, I oversample 90% of the other tets by a factor of 1000.

The application is for a volume render that runs inside of RenderMan. I've
written a shader that use CallModule to compute the color and opacity at the
requested point. I also have a slice shader. You pass to the the shader the
dataset name and a colormap and dx computes the color. Pretty cool.

However, the shaders are slower than molasses on tetrehedral grids (not bad on
regular). For the slice shader, it's just doing a Map on the point generated by
RenderMan. Since each point is taken in isolation (there is no assumption of the
point being on a plane, the "surface" being shaded could be a cylinder for
instance), you pay a massive overhead for the cell search. I think the best
thing for me is to write a new module that would cache the point and start the
search for the next cell from the last cell.


Ideas?

Mark

--
   Mark Bolstad
   Lockheed-Martin Services Inc.
   [EMAIL PROTECTED]
   (919)-541-3604

Reply via email to