Ray tracing speed-up techniques using an MIMD architecture

Abstract

Since its appearance in computer graphics research, ray tracing has attracted considerable attention, especially with regard to speeding up the algorithm. This paper presents an acceleration scheme based on utilizing parallel processing techniques. It is shown how an existing ray tracing acceleration technique - Amanatides' voxel transversal algorithm - implemented on an MIMD processor architecture accelerates performance drastically. Aside from being used for speeding up purposes, the voxel structure also serves - and this is at least just as important - as a means for minimizing communication overhead regarding the distributed database. Additional performance-improving features within our dynamic load-balanced system are described also.