This paper presents the results of an exploratory experiment investigating the use of program metrics to predict reachability graph size, reachability graph generation time, and deadlock check times for concurrent Ada programs. In general, reachability analysis is intractable, so it is important to be able to predict whether or not this analysis technique is feasible given a specific program to be analyzed. We briefly introduce the program representations used to perform the deadlock reachability analysis and describe a set of metrics on those representations. We present empirical results that quantify the value of each metric as a predictor of the reachability graph size and the practicality of deadlock reachability analysis.
A. T. Chamillard