Parallel Ray Tracing with 5D Adaptive Subdivision

We present strategies for parallelising ray tracing based on 5D adaptive subdivision. Our goals are to obtain good speed-up and to efficiently balance the load between the processors while minimising the required memory per processor inherently large in 5D subdivision. First, loosely coupled strategies are presented, which are ideal for implementation on clusters of workstations, the most commonly used form of parallel processing nowadays. Then we consider a tightly coupled algorithm ideal for multiprocessors with fast interconnection network or shared memory. Finally, results on a cluster of workstations are presented and discussed.

• Type: Conference
• Year: 2001
• Where: WSCG
• Authors: G. Simiakakis, Theoharis Theoharis, A. M. Day
• Keywords: Adaptive Subdivision, Fast Interconnection Network, Required Memory, WSCG 2001