A Region-based Approach for Sparse Parallel Computing

Bradford L. Chamberlain
E Christopher Lewis
Lawrence Snyder

University of Washington Technical Report UW-CSE-98-11-01
(presented as a poster at LCPC'96)

Abstract: This paper introduces a technique for parallel sparse computation by extending the array-language concept of regions--regular programmer-specified index sets used for specifying array computations. We introduce the notion of sparse regions which can represent an arbitrary set of indices. Sparse regions inherit the benefits of regular regions, including conciseness, a direct encapsulation of parallelism, and support for language performance models that highlight parallel overheads. We show that region-based array languages can benefit from the use of sparse regions, both in terms of the semantic richness available to the programmer and the execution times of the resulting program. We also demonstrate that regions result in efficient implementations as compared to array-based approaches, due to their role in amortizing sparse overheads and enabling optimizations.

postscript | PDF