Efficient parallel recursive voxelization for SGI Challenge multi-processor system

Recursive algorithms are generally not straightforward to be processed in parallel, particularly when an octree is involved. Some researchers have tried to process an octree in parallel, but this solution is restricted to recursive subdivision methods that guarantee surface containment in an octant. We show a whole class of recursive subdivision algorithms which does not fall into this category. Our approach solves this problem, accepting any kind of spatial recursive subdivision algorithm, by accessing and creating the octree serially, while the subdivision process is done in parallel. Thanks to our efficient octree generation algorithm, its access and creation times are meaningless in comparison to the subdivision process. We further show that our solution gives excellent results to the problem of surface voxelization using spatial recursive subdivision. The method can be also successfully applied to other computer intensive problems using spatial recursive subdivision.