Abstract
The objective of this thesis is to design and implement the ray tracing algorithm on the distributed-memory message- passing parallel multicomputer. Two-phased parallel ray tracing is proposed to avoid the load imbalance and oscillating problems in the dynamic load balancing method. On the assumption that the first phase is already implemented on the sequential or parallel computer and the database is optimally distributed among the nodes of the parallel computer, the second phase is implemented on the general- purpose multicomputer, the NCUBE hypercube parallel computer. The object space of an image scene is first organized in the candidate object grid, which is distributed among the processors of the NCUBE parallel multicomputer. Then each processor detects simultaneously the intersections of the surfaces of each object and a ray. Processors communicate by means of message-passing to locate and request external objects. This scheme allows a database of large size to be supported on multicomputers with limited distributed memory. Implementation results are presented to illustrate these features. For future research, the cache scheme and the multithreaded system can be incorporated into the current implementation. At the same time the first phase for static load balancing should be studied for the better performance.
Kwon, Young-Jun (1993). An implementation of a parallel graphics algorithm on a distributed-memory message-passing parallel computer. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1993 -THESIS -K98.