Symbolic techniques for formally verifying industrial systems

Abstract

The design of correct computer systems is extremely difficult. However, it is also a very important task. Such systems are frequently used in applications where failures can have catastrophic consequences, or cause significant financial losses. Simulation and testing are the most widely used verification techniques, but they can only show the presence of errors and cannot demonstrate correctness. Until lately formal methods were too expensive to be used in industrial problems, but recent research has made it possible to apply formal techniques to the verification of complex real-world systems. Symbolic model checking is an example of such a technique that has been successful in verifying large finite-state systems. It has also been extended to produce timing and performance information. These properties are extremely important in the design of high-performance systems and time-critical applications. A more detailed analysis of a model is possible using these extensions than by simply determining whether a property is satisfied or not. We present algorithms that determine the exact bounds on the delay between two specified events and the number of occurrences of another event in all such intervals. To demonstrate how our method works, we present two complex examples: the verification of the Futurebus+ cache coherence protocol and the timing analysis of the PCI local bus. These results show the usefulness of symbolic model checking in analyzing modern industrial designs.