Dynamic reordering in a breadth-first manipulation based BDD package: challenges and solutions

Abstract

The breadth-first manipulation technique has proven effective in dealing with very large sized BDDs. However, until now the lack of dynamic variable reordering has remained an obstacle in its acceptance. The goal of the work is to provide efficient techniques to address this issue. After identifying the problems with implementing variable swapping (the core operation in dynamic reordering) in breadth-first based packages, the authors propose techniques to handle the computational and memory overheads. They feel that combining dynamic reordering with the powerful manipulation algorithms of a breadth-first based scheme can significantly enhance the performance of BDD based algorithms. The efficiency of the proposed techniques is demonstrated on a range of examples.