Minimizing ROBDD size of incompletely specified multiple output functions

The size of ROBDD is essential in many applications. In this paper we propose a heuristic that minimizes ROBDD size for incompletely specified multiple output functions. Our technique involves giving a new interpretation of ROBDDs with an extra variable, Z for incompletely specified functions. This new interpretation allows us to assign don't cares to either the on-set or the off-set yielding an efficient minimization of ROBDD size. The extra variable Z carries the information of the don't cares and preserves the flexibility of choosing the best cover for a given incompletely specified function. Starting from the top level of an ROBDD, we minimize greedily the number of nodes at every level by assigning as few don't cares as possible to either the on-set or the off-set. We propagate the remaining unused don't cares to lower levels of an ROBDD by moving the extra variable downwards. We performed three sets of experiments. The results are very encouraging.<
>