An optimization technique for ordered (binary) decision diagrams

Abstract

The minimum-cost ordered (binary) decision diagram (OBDD) (also a reduced OBDD or ROBDD) is a canonical representation for a logic function, given an ordering on its variables (R. Bryant, 1986). A new optimization technique is presented for suboptimal synthesis of ODDs of complete as well as partial multiple-output Boolean functions. The method is based on iterative decomposition. The central notion in this process is that of subfunctions, whereas in ODDs there are decision nodes. There is, however, 1:1 mapping between them: a level of decision nodes in the ODD corresponds to a set of subfunctions recognized in a corresponding decomposition step. The technique is computationally effective and deals with incomplete functions frequently used in practice. A small synthesis example is given to introduce a new technique for ROBDDs. The results and some experience with the optimization program are described.<>