An Improved Reversible Circuit Synthesis Approach using Clustering of ESOP Cubes

Abstract

The problem of reversible logic synthesis has drawn the attention of many researchers over the last two decades with growing emphasis on low-power design. Among the various synthesis approaches that have been reported, the ones based on compact circuit representations like Binary Decision Diagrams (BDD) and Exclusive-or Sum-Of-Products (ESOP) are interesting in the sense that they can handle large circuits with more than 100 inputs. The drawback of these approaches, however, is that the generated netlists are sub-optimal, and there is lot of scope for optimizing them. One of the best methods in this regard is an approach, where the ESOP cubes are grouped into sublists based on sharing among more than one outputs. In the work reported in this article, in contrast, an approach based on clustering the ESOP cubes based on their similarity with respect to input variables is presented, along with a technique to map each of the clusters into reversible gate netlists. This approach results in a significant reduction in quantum cost of the final netlist, but requires one additional garbage line. Experimental results on a number of reversible circuit benchmarks have been presented in support of the claim and also demonstrate that the method is very fast.