Data structures in ray-tracing

[Mika]

Hello,

Given that I'm working with more scientific ray-tracing, I realized that some of the conventions used in real time rendering don't apply the kind of ray-tracing I'm working with. For example, sometimes I need to save the path the ray took, but on the other hand I don't expect there would be many objects on the scene. The problem I'm now facing is that I don't actually have sufficient background of modern computer architecture to understand what kind of data structure would be most efficient in handling ray paths where there are different amount of nodes. Would anyone have any suggestions or a book to help me with this issue?

Also, most of the surfaces I'm working with are spherical, aspherical or splines. By testing, I found that the spline is the most useful for the application, but it is also the most computationally expensive. Does anyone have some spare time to help me to find the correct methods to tackle with this? With some special types of splines, I algebraically reduced the amount of calculations to smallest I could find, but somehow I suspect writing evaluation routines that way would not provide the best outcome.