Accelerated 5D Ray Tree construction on the GPU

Abstract

Ray tracing random rays has been a challenge. Due to their reduced coherence, the normal methods of acceleration like packet tracing that make use of coherence of the rays do not work well. These random rays are encountered in global illumination methods. 5D ray classification, first introduced by Arvo and Kirk [AK87], can classify these rays into coherent groups. We introduce a method that builds a hierarchical structure identifying coherence in random rays very quickly using the increased processing power of the GPU.