Approximating the safely reusable set of learned facts

Abstract

Despite many advances, today’s software model checkers and extended static checkers still do not scale well to large code bases when verifying properties that depend on complex interprocedural flow of data. An obvious approach to improve performance is to exploit software structure. Although a tremendous amount of work has been done on exploiting structure at various levels of granularity, the fine-grained shared structure among multiple verification conditions has been largely ignored. In this paper, we formalize the notion of shared structure among verification conditions and propose a novel and efficient approach to exploit this sharing by safely reusing facts learned while checking one verification condition to help solve the others. Experimental results show that this approach can improve the performance of verification, even on path- and context-sensitive and dataflow-intensive properties.