A parallel implementation of bidirectional ray tracing on transputer-based system

Abstract

This paper describes a parallel processing scheme for bidirectional ray tracing which improves the reality of the image drastically. It is organized from two stages, i.e. light ray tracing and viewing ray tracing. In light ray tracing, pairs of a light source and an element of objects are distributed to each processor dynamically. In viewing ray tracing, we use a screen subdivision algorithm. Our system consists of 86 transputers. All of the scene data are placed on shared memories, and only the required parts of them can be transferred to the cache memory on each processor by means of DMA on-the-Hy. We also show some experimental results.