A data partitioning algorithm for distributed memory compilation

Abstract

This paper proposes a compiler strategy for mapping FORTRAN programs onto distributed memory computers. Once the available parallelism has been identified, the minimisation of different costs will suggest different data and computation partitions. This is further complicated, as the effectiveness of the partition will depend on later compiler optimisations. For this reason, partitioning is at the crux point of compilation and this paper describes an automatic data partition algorithm which is based on the analysis of four distinct factors. By determining the relative merit of each form of analysis, a data partitioning decision is made which is part of an overall compilation strategy. The strategy is applied to a real non-trivial program on a 32 cell KSR-1 where the performance is comparable to that of hand-coded techniques.