> I am not sure about this, but I assume that it is the calculation of the
> lightfall and shadows of the object elements that takes all the time?

I'm sure if you search around you can find data on these NGs somewhere, 
but basically the thing that typically takes most of the time in 
raytracing is figuring out which object(s) a ray actually intersects.

(They recon ray intersection tests typically make up 80% of the 
calculation - obviously depending on what you're drawing.)

This is probably *especially* true with isosurfaces. Gotta do a whole 
adaptive sampling process for each ray intersection there. (Plus extra 
calculations to find the surface normal.)

Once you have the ray intersection coordinates and the surface normal, 
actually figuring out the lighting is a tiny little calculation. (Except 
that if you want shadows, gotta shoot *more* rays... And if you have 
area lights, gotta shoot *loads* more rays...)

This is why POV-Ray [currently] has bounding boxes, vista buffers, light 
buffers, etc. It's all the try to reduce the number of ray intersection 
tests...

[I understand the new beta might be using kD trees or something... not 
finalised yet.]

Post a reply to this message