Bounded Model Checking (BMC) encodes a model checking problem in the
propositional logic. Diagnosing the resulting formula as satisfiable
provides a counterexample. While surprisingly efficient for many
complex systems, in general BMC still fails to be complete and is
a method of falsification rather then validation. The major obstacle is
satisfiability testing (SAT). The talk introduces a selective search to
the standard DLL SAT algorithm, allowing to profit from several optimization
techniques proposed for non-symbolic methods. Partial-order reductions
are shown as an example of selective search. Preliminary experimental
results confirm that the selective search can significantly improve
the effectiveness of BMC.
|