An empirical evaluation of three methods for deadlock analysis of Ada tasking programs

Abstract

Static analysis of Ada tasking programs has been hindered by the well known state explosion problem that arises in the verification of concurrent systems. Many different techniques have been proposed to combat this state explosion. All proposed methods excel on certain kinds of systems, but there is little empirical data comparing the performance of the methods. In this paper, we select one representative from each of three very different approaches to the state explosion problem: partial-orders (representing state-space reductions), symbolic model checking (representing OBDD-based approaches), and inequality necessary conditions (representing integer programming-based approaches). We apply the methods to several scalable concurrency examples from the literature and to one real Ada tasking program. The results of these experiments are presented and their significance is discussed.