Demonstration of automatic data partitioning techniques for parallelizing compilers on multicomputers

An approach to the problem of automatic data partitioning is introduced. The notion of constraints on data distribution is presented, and it is shown how, based on performance considerations, a compiler identifies constraints to be imposed on the distribution of various data structures. These constraints are then combined by the compiler to obtain a complete and consistent picture of the data distribution scheme, one that offers good performance in terms of the overall execution time. Results of a study performed on Fortran programs taken from the Linpack and Eispack libraries and the Perfect Benchmarks to determine the applicability of the approach to real programs are presented. The results are very encouraging, and demonstrate the feasibility of automatic data partitioning for programs with regular computations that may be statically analyzed, which covers an extremely significant class of scientific application programs