Mining Complex Boolean Expressions for Sequential Equivalence Checking

We propose a novel technique to mine powerful and generalized boolean relations among flip-flops in a sequential circuit for sequential equivalence checking. In contrast to traditional learning methods, our mining algorithm can detect inductive invariants as well as illegal state cubes. These invariants can be arbitrary boolean expressions and can thus prune a large don't care space during equivalence checking. Experimental results demonstrate that these general invariants can be very effective for sequential equivalence checking of circuits with no or very few equivalent signals between them, with low computational costs.