DESIGN OF PARALLEL ALGORITHMS FOR A DISTRIBUTED MEMORY HYPERCUBE

In order to use a multiprocessor effectively it is necessary to develop efficient techniques for the assignment of the computations to the processors available. We must previously be capable of identifying all the computations that can be executed in parallel. In this paper we present a procedure for the efficient programming of a distributed memory, message-passing hypercube multicomputer. This procedure is applied to a sequential program and results in a parallel program which can be expressed in a parallel language. We also present a language with these characteristics called ACLAN (Array C LANguage), as well as the software package ACLE (Array C Language Emulator) for the simulation of ACLAN programs in sequential computers. We have satisfactorily applied the parallelization procedure to a large number of algorithms in the fields of matrix algebra and image processing. We present the results obtained.