Controlling the memory during manipulation of word-level decision diagrams

Abstract

Decision diagrams (DDs) are efficient data structures that are frequently used for formal verification, test or synthesis of circuits and systems. The main drawback of the data structure is the potential memory blow-up caused by certain functions. Therefore streaming has been proposed as a technique to efficiently evaluate operations on binary DDs (BDDs). The maximal memory usage can be controlled and therefore calculations can be carried out with a small amount of memory. In this paper we show how streaming can be extended to word-level decision diagrams (WLDDs) and for which type of diagrams streaming is applicable. A detailed description of streaming for multi-terminal BDDs (MTBDDs) is given. Experiments show the efficiency and the small memory needs for operations on MTBDDs.