Abstract

One of the most popular interconnection networks for parallel computers today is the binary n-cube, also called the hypercube, interconnection network. The attractiveness of the hypercube topology is its small diameter, simple and efficient message routing algorithms, strong connectivity, regularity, symmetry, and fault-tolerance. Unfortunately, conventional VLSI technology seems to be reaching its fundamental physical limits and is, therefore, unable to implement large-scale interconnection networks based on the hypercube topology (or any other topology for that matter) that provide high communication bandwidth, short latency, and reasonable power requirements[1].

© 1993 Optical Society of America