An analysis of ATPG and SAT algorithms for formal verification

Abstract

We analyze the performance of satisfiability (SAT) and Automatic Test Pattern Generation (ATPG) algorithms in two state-of-the-art solvers. The goal is to best understand how features of each solver are suited for hardware verification. For ATPG, we analyze depth-first and breadth-first decision orderings and effects of two weighting heuristics in the decision ordering, and also study the effect of randomization of decisions. Features of ATPG and SAT that affect their robustness and flexibility on real circuits are studied, and the two solvers are compared on 24 industrial circuits. We further analyze the results to identify the strengths and shortcomings of each solver. This will enable incorporation of features from each solver in order to optimize performance, since they both operate on the same principles.