Powerful and feasible processor interconnections with an evaluation of their communications capabilities

Abstract

Scalable networks with very good topological properties are often impossible to build because of their prohibitively high wiring complexity. Such a network is the generalized hypercube (GH). It supports full-connectivity of all its nodes in each dimension and is characterized by outstanding topological properties. We propose a new class of scalable interprocessor connections, namely HOWs (Highly-Overlapping Windows), capable of lower complexity than GHs and comparable performance. HOWs are obtained from GHs by removing edges to produce systems of lower wiring complexity. They contain numerous highly-overlapping GHs of smaller size. The classical GH belongs to this new class of interconnections. We demonstrate that 2-DHOWs perform much better than binary hypercubes for important communications patterns.