Managing space for finite-state verification

Abstract

Finite-state verification (FSV) techniques attempt to prove properties about a model of a system by examining all possible behaviors of that model. This approach suffers from the state-explosion problem, where the size of the model or the analysis costs may be exponentially large with respect to the size of the system. Using symbolic data structures to represent subsets of the state space has been shown to usually be an effective optimization approach for hardware verification. The value for software verification, however, is still unclear. In this paper, we investigate applying two symbolic data structures, Binary Decision Diagrams (BDDs) and Zero-suppressed Binary Decision Diagrams (ZDDs), in two FSV tools, LTSA and FLAVERS. We describe an experiment showing that these two symbolic approaches can improve the performance of both FSV tools and are more efficient than two other algorithms that store the state space explicitly. Moreover, the ZDD-based approach often runs faster and can handle larger systems than the BDD-based approach.