If primitives are stored in multiple nodes then that makes adding and removing nodes and in general changing the structure harder. Also maybe if primitives are in disjoint sets we can implement parallelism on each disjoint set without doing redundant work.
If primitives are stored in multiple nodes then that makes adding and removing nodes and in general changing the structure harder. Also maybe if primitives are in disjoint sets we can implement parallelism on each disjoint set without doing redundant work.