Builds a tree over boxes, as buildSubtree, much faster and not as good: a linear BVH
(Karras, 2012). The leaves are sorted along the Morton curve of their centres (10 bits per axis
over the centres' bounds, by a radix sort), and each node splits its leaves where their codes
first differ, in the middle where they are the same. Linear in the leaves: for the culling and
the picking of a model being loaded, BoxTree.rebuild making a better tree once it is.
Builds a tree over boxes, as buildSubtree, much faster and not as good: a linear BVH (Karras, 2012). The leaves are sorted along the Morton curve of their centres (10 bits per axis over the centres' bounds, by a radix sort), and each node splits its leaves where their codes first differ, in the middle where they are the same. Linear in the leaves: for the culling and the picking of a model being loaded,
BoxTree.rebuildmaking a better tree once it is.