A hypergraph-based interconnection network for large multicomputers

L. M. Mackenzie, M. Ould-Khaoua, R. J. Sutherland, T. Kelly

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The bandwidth constraints placed on architectures by implementation technology are a critical determining factor in how well the theoretical properties of a particular topology can be exploited. However, great care must be taken in deciding which constraint is applicable in a particular technology and due attention given to examining whether a topology may, by its nature, be able to exploit partitioning and interconnection design to alleviate the effects of that constraint. This paper reviews current thinking on these matters, and introduces a new hypermesh architecture called COBRA, together with an associated interconnection scheme which permits it to reduce substantially the effects of conventional bandwidth limitations.

Original languageEnglish
Title of host publicationParallel Processing
Subtitle of host publicationCONPAR 1992 ─ VAPP V - 2nd Joint International Conference on Vector and Parallel Processing, Proceedings
EditorsLuc Bouge, Michel Cosnard, Yves Robert, Denis Trystram
PublisherSpringer-Verlag
Pages837-838
Number of pages2
ISBN (Print)9783540558958
DOIs
Publication statusPublished - 1992
Event2nd Joint International Conference on Vector and Parallel Processing, CONPAR 1992 - Lyon, France
Duration: Sep 1 1992Sep 4 1992

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume634 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2nd Joint International Conference on Vector and Parallel Processing, CONPAR 1992
Country/TerritoryFrance
CityLyon
Period9/1/929/4/92

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this