Exploiting isomorphism for compaction and faster simulation of binary decision diagrams

Abstract

We present two techniques for compaction of ROBDDs. The first technique extracts isomorphic subtrees from a characteristic function ROBDD (cfBDD), replaces them by multi-output nodes, and stores the extracted subtrees as MTBDDs. The second technique searches pre-defined topological structures (signatures) within the cfBDD and replaces them by multi-output nodes. While both approaches are able to extract isomporphic subtrees in the cfBDD, the signature scanning approach gives a significantly better compression and reduces the simulation time as compared to cfBDD simulation, which shows that is possible to compress BDDs and yet simulate them faster.