Parallel Computing: Perspectives and Prospects

Abstract

The trend toward parallel processing is driven by both scientific and economic factors. At the high end of the performance range, parallelism is the only answer to the fundamental limits on computational speed that are being rapidly approached by today’s supercomputers. In the lower and middle performance ranges, parallel architectures assembled from commodity parts, such as microprocessors and memory chips, provide the most cost effective systems. These trends are mutually supportive, as the lessons learned and software developed are transferred across systems.