Currently the BVHRebuilder will create a dummy internal node when in fact only a single leaf would suffice. The SplitBVHBuilder will do the same.
The traversal algorithms (the generic ones as well as the optimzed speculative kernel one) have no real dependency for this.
Redesigning the BVHRebuilder will be the bulk of the work. Simplifying the SplitBVHBuilder will also be necessary.
Currently the BVHRebuilder will create a dummy internal node when in fact only a single leaf would suffice. The SplitBVHBuilder will do the same. The traversal algorithms (the generic ones as well as the optimzed speculative kernel one) have no real dependency for this. Redesigning the BVHRebuilder will be the bulk of the work. Simplifying the SplitBVHBuilder will also be necessary.