Dynamic proximity clouds on the GPU

Abstract

Ray tracing is a widely used technique in rendering realistic scenes in Computer Graphics. Its main drawback has been that it is time consuming, requiring the rendering to finish from hours to sometimes days. For decades, the goal has been to speed up the processing of these scenes. Two popular grid traversal techniques have emerged: (a) Three Dimensional Digital Differential Analyzer (3DDDA) and (b) the Proximity Cloud (PC), which is a variation of 3DDDA. Both of these techniques try to limit the number of collision tests, which can be the most time consuming part of the algorithm. While both techniques allow impressive speedups, large dynamical varying scenes topology still challenge the real time rendering process. These techniques are optimal on static scenes, but object movement forces recalculation of the scene. This is a problem when using CPUs because parallelization is not easily available. Running these on GPUs, however, allows for parallelization. Apart from briefly summarizing some of our previous results from Ryan and Semwal (Proceedings of the world congress on engineering and computer science 2014, San Francisco, pp. 376–381 [1]), we also look to answer some of the more relevant questions about ray tracing, and what future holds for this area.