PERT A PIPELINED ENGINE FOR RAY TRACING GRAPHICS
Pradeep Chilka · Summit (Simon Fraser University) · 1985
Ray tracing techniques for image rendering have produced some of the most realistic images to date. Ray tracing, however, is computationally expensive because of the floating point calculation involved in ray-object intersection and the number of such intersections that must be performed t o render an image realistically Conventional mini-computers take anywhere between an hour and several days to render a single image of moderate complexity In this thesis, we propose a pipelined machine. PERT, which according to our simulation results, shows a substantial reduction in the rendering time. The key features of PERT are: a) the use of bounding volumes and hierarchical data organization to reduce the number of ray-object intersections, b) a 3-processor pipeline that executes a 3-task ray tracing algorithm, c) microcoded, custom designed. VLSl processors in each stage of the pipeline, and d) extensibility t o a multi-PERT architecture that consists of several PERTs working in parallel.