first allow me to thank you for this very well-written piece of software, it is a true gem
I have one question and one suggestion.
the question is: where are you planning to take Embree? are we going to see future versions of Embree or 1.0 will be the last? and will these versions (if planned) implement new features or more optimizations?
and the suggestion is: since CPU easily overtakes the GPU in biased rendering methods, I think it might be very useful to include SSE optimized algorithms for Octrees (for irradiance caching) and Knn finding (for photon mapping). What do you think?
There is no official roadmap for Embree, because it is still considered a research project. We are planning to evaluate the feedback we get for the first release and we'll then decide on a case-by-case basis if and when to add new features and optimizations.
The primary goal for Embree is to provide highly optimized implementations of the performance critical parts of a photo-realistic renderer. I agree with you that octrees and Knn search clearly fall into that category.