GPU Path Tracing

Abstract

The goal of this work is to verify the possibility to utilize GPU for global illumination computations in a commercial software environment and explore an efficient way to do it. Path tracing with BVH as the acceleration data structure was implemented on GPU using CUDA successfully. It was arranged as a pipelined structure which supported mulitple texture types and light source types. And it also proved to be efficient in execution and compact in memory. A light tree structure was introduced in this work as well, which grouped light sources with an affecting radius hierarchically and enabled the path tracer to handle massive light sources efficiently. Testings and analyses were also made for different configurations to utilize the light tree.