Ada on a hypercube

Abstract

The widespread use of parallel machines, and hypercubes in particular, is being held back by the lack of high-order parallel programming languages. In this paper we discuss the issues involved in establishing an existing language that supports parallel processing, Ada, on a hypercube multiprocessor. The majority of the paper addresses the requirements and implementation of the run-time system, which is the key to establishing any parallel language. First, the requirements of the run-time system for Ada are described from a machine-independent point of view. Next, the approach taken toward implementing this system on a hypercube is discussed, with considerations given for language-level program partitioning and interprocessor communication performance. Finally, the status of our current implementation is discussed and some concluding remarks are made about parallel languages in general, based on our experiences.