Efficient algorithms for data distribution on distributed memory multicomputers

Abstract

Data distribution has been one of the most important research topics in parallelizing compilers for distributed memory parallel computers. In this paper, we show that data re-distribution is necessary for executing a sequence of Do-loops if the communication cost due to perform this sequence of Do-loops is larger than a threshold value. Based on this observation, we propose efficient algorithms which can determine effective data distribution schema for executing a sequence of Do-loops with a general structure. Our result contributes towards automatic compilation of sequential programs to message-passing version programs running on distributed memory parallel computers.