Ray tracing and binary tree computations using PVM

Abstract

The authors investigate the design, analysis, and implementation issues of algorithms for two computational problems-namely, ray tracing and bottom-up binary tree computation-in a distributed setting. The implementations have been performed on a cluster of workstations running on parallel virtual machine (PVM). Ray tracing algorithms for rendering computer synthesized images are computationally expensive due to the very long computing time required to calculate the intersection of each ray with all objects. The authors propose a variant of a recent static load balancing technique proposed by T. Priol and K. Boatouch (1989), where the sampling is based on partitioning the object space. This usually results in poor processor efficiency. The approach to this is to partition the image instead and use an efficient scheduling technique to allocate work to individual processors. The advantage of the distributed algorithm is that it can be used with any sequential algorithm for ray tracing. The other problem addressed is distributed binary tree computations. The tree contraction problem in a distributed setting is studied. >