Downloading node programs/data into hypercubes

In a multiprocessor environment, I/O (loading/unloading of node's or processor's program/data) time has significant impact on the overall execution time of a program. To minimize the I/O time one has to carefully route the data across the interconnecting network. In this paper different data loading strategies are discussed for hypercube multiprocessors with a host processor. A new loading strategy is proposed in which both the host and nodes are kept busy simultaneously. The effectiveness of this strategy is demonstrated by comparing with the existing methods using maximum finding and sorting algorithms. © 1991 Elsevier Science Publishers B.V. All rights reserved.