This paper presents experimental results on the performance effect of using symbolic simulation with SAT-based reparametrization within the Counterexample Guided Abstraction Refinement framework. Abstraction refinement has been applied successfully to prove safety properties of large industrial circuits. However, all existing abstraction refinement frameworks simply use SAT-based Bounded Model Checking (BMC) to refute the property. The model used for the BMC instance is not abstracted, and thus is susceptible to the state space explosion problem. We address this issue by using a symbolic simulator with a SAT-based reparametrization algorithm as a replacement for BMC within the abstraction refinement framework. The reparametrization is performed as soon as the equations maintained by the symbolic simulator become too large. We discuss the quality of the refinement information that is extracted from the symbolic simulator.

19 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu