ADDT: Automatic Data Distribution Tool for Porting Programs to PVM

In this paper we present a method to obtain a set of candidate distributions for a program fragment. These candidate distributions can then be used to obtain a dynamic data distribution for the whole program using the method outlined in this paper. Our algorithms first find an inter-axis alignment for the arrays in the distribution block. Using this axis alignment a data distribution is obtained for these arrays. The algorithms presented here aim to obtain an alignment-distribution that minimizes data movement between processors with minimal loss of parallelism. We have presented experimental results obtained by executing these algorithms on several programs. 1 Introduction Parallel processing is increasingly being accepted as a viable approach to solve computationally intensive applications in the sciences and in engineering. Research in parallel processing systems over the last decade has resulted in the development of systems with widely This research is supported by the Boeing ..