2000pdcs William Hui, Steve MacDonald, Jonathan Schaeffer and Duane Szafron, Visualizing Object and Method Granularity for Program Parallelization, to appear in Proceedings of Parallel and Distributed Computing and Systems (PDCS 2000), November 2000, Las Vegas U.S.A, pp. 286-291. abstract or pdf.

There are increasing demands for more computing power. Parallel hardware is now commonplace and could be a cost-effective solution. However, to many developers, parallel programming is not a user-friendly task. At the same time, many programmers are turning to object-oriented techniques. Unfortunately, parallel computing with objects introduces many new problems. Tools are needed to help programmers convert their sequential object-oriented programs into parallel ones. This paper introduces Revy, a platform-neutral, easy-to-use, object/method granularity visualization system that assists parallel programmers in transforming their sequential object-oriented programs into parallel ones. Revy allows users to view and inspect the object communication patterns of their sequential applications and it serves as a profiling system that helps them identify the high-granularity objects/methods as candidates for parallel execution. This paper describes the requirements, architecture and implementation of Revy and illustrates the ideas with a case study.