SBDD variable reordering based on probabilistic and evolutionary algorithms

Modern CAD tools must represent large Boolean functions compactly in order to obtain reasonable runtimes for synthesis and verification. 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. A common heuristic for the variable ordering problem is to group variables together that have similar characteristics. We use this heuristic to formulate a technique for the reordering problem using probability based metrics. Our results indicate that this technique outperforms sifting with comparable runtimes. Furthermore, the method is robust in that the final results independent of the initial structure of the SBDD