Inferring simple solutions to recursion-free horn clauses via sampling

Recursion-free Horn-clause constraints have received much recent attention in the verification community. It extends Craig interpolation, and is proposed as a unifying formalism for expressing abstraction refinement. In abstraction refinement, it is often desirable to infer “simple” refinements, and researchers have studied techniques for inferring simple Craig interpolants. Drawing on the line of work, this paper presents a technique for inferring simple solutions to recursion-free Hornclause constraints. Our contribution is a constraint solving algorithm that lazily samples fragments of the given constraints whose solution spaces are used to form a simple solution for the whole. We have implemented a prototype of the constraint solving algorithm in a verification tool, and have confirmed that it is able to infer simple solutions that aid the verification process.