Variable Reordering for Shared Binary Decision Diagrams Using Output Probabilities

Abstract

The Shared Binary Decision Diagram (SBDD) with negative edge attributes can represent many functions in a compact form if a proper variable ordering is used. In this work we describe a technique for reordering the variables in an SBDD to reduce the size of the data structure. We use a heuristic to formulate a technique for the reordering problem based on probability metrics. 1 Introduction Many approaches for BDD variable reordering have been investigated. Techniques that depend on variable interrelationships such as symmetry have been applied to both variable ordering and reordering approaches [2, 4, 7, 5, 6]. Recently, it has been observed that the robustness of OBDD algorithms is a very important aspect [3], but the dynamic variable ordering algorithms above largely depend on the chosen starting point, since they are mainly greedy heuristics. In this paper we describe a stable variable ordering heuristic that makes use of properties of the function only, but is not dependent on t...