Donald Bren School of Information and Computer Sciences

An approach to statistically analyzing a concurrent program not suited for analysis is described. The program is reengineered to reduce the complexity of the concurrency-related activities, thereby reducing the concurrency state space. The key to the reengineering process is a metric set that characterizes program task interaction complexity and provides guidance for restructuring. An initial version of a metric set is proposed and applied to two examples to demonstrate the utility of the reengineering-for-analysis proces. The reengineering has potential benefits apart from supporting analyzability, following the dictum that if it is hard to analyze, it is hard to understand and maintain.