Compositional Analysis for Weak Stubborn Sets

Partial order reduction methods rely on expanding a state space only partially, exploring representatives of sets of executions of a system. The methods differ at least in three respects:the set of properties that are preserved, the notions of interaction between transitions, and the methods of selecting the representative transitions. We explore an existing but less studied method of weak stubborn sets in the context of process-algebraic parallel composition. The theory of stubborn set methods is based on global condition son executions. In practice, these conditions are guaranteed by using static dependency information in the form of dependency and causality relations. We propose a compositional approach for these conditions. We extract dependency information from the component processes by analysing the components in detail and then define compositional rules for system-level dependency information. We use novel localised conditions of dependency that the weak stubborn set method can make use of. We carry out some experiments in the context of FDR, to explore the relative merits of the method compared to the more well-known version of stubborn sets with and without compositional analysis. We do this for both a deadlock preserving reduction and a version that preserves the failures/divergences semantics of FDR.