I guess ray tracing can start as very uniform calculation. But depending on the nature of the surfaces, it may become less parallelizable (different rays wander off and some may "disappear" ver early on). Is this accurate? If so, how does GPUs handle this efficiently with the wide SIMDs?
Is it possible to mix ray casting and rasterization? Or is it preferable to stick to one technique for a particular application?
Do some rasterization pipelines in things like video games that require quick rendering every use raytracing to render reflective objects, or things that are hard to do with rasterization, or is this only done in a fully raytraced environment such as RTX?
Do reflections cause problems for organized trees if they are optimized to get rays from the origin?