An Improvement of Quasi-weakest Precondition Inference using Fast MCS Enumeration

We previously proposed a method to infer preconditions that are the weakest in a combination of predicates (quasi-weakest preconditions). This method, however, had a problem in performance due to the high cost of minimal-unsatisfiable-core (or MUC) enumeration. MUC enumeration is usually realized by using enumeration of minimal correction subsets (or MCSes) as intermediate solutions, and the MCS enumeration forms large part of MUC enumeration time. In this paper, we propose three fast algorithms for quasi-weakest precondition inference, based on two properties of MCS enumeration: (1) in our setting the size of MCSes can be fixed, and (2) they can be efficently enumerated if some of the MUCs are obtained in advance. Our performance evaluation shows all of these three algorithms were superior compared to a conventional one, with a maximum speedup of 10.7 times. We report the result of the evaluation, and discuss pros and cons between the three.