|
I am adding 3D affine transformations to my ray tracer, and it seems all the literature on the web recommend inverse transforming the ray instead of transforming the 3D objects. Why is that? There are a lot more rays than objects in a typical ray tracing scene, so wouldn't it be more efficient to transform the objects themselves? |
|||
|
|
|
In a typical raytracer, you are not just raytracing triangles and vertices, but various implicit surfaces. Each has their own associated hit test. In general, it's a lot easier to just inverse transform the ray than to modify every specific hit test. |
|||
|
|
|
The number of vertices that would require transforming may be higher than the number of rays. I haven't written a raytracer so I wouldn't know, but all I'm saying is that the number of distinct objects is irrelevant- only the number of verts. |
|||
|
|
