A Roadmap for a Type Architecture Based Parallel Programming Language

Abstract

Ever since the end of the era of single processor performance improvement, we observe proliferation of multi and many-core architectures in almost all spheres of computing. Tablets, desktop PCs, workstation clusters, and supercomputers are ripe with multi-core CPUs and/or accelerators. Although there are considerable architectural heterogeneity and scale differences present among these machine architectures, there is also significant commonality in their multi-core building blocks. This situation revives the interest for possibility of parallel programming paradigms that are both efficient and portable across environments. We have been investigating such a programming paradigm for over a year and half now. Two key aspects of our work are development of a common machine abstraction for diverse hardware platforms and a clear separation of different concerns embodied in parallel programming. This is our first paper describing the philosophy, abstraction mechanism, programming model, and early results of our ongoing project.