Approximative Representation of Boolean Functions by Size Controllable Robdd's

ROBDD's((2]) are a very popular datastructure for the representation and manipulation of boolean functions. They have tractable sizes for many boolean functions and come up with eecient algorithms for boolean operations. In the worst case however, size and time complexity grows exponentially when performing a polynomial number of operations. However, there are applications where an approximate knowledge about a boolean function like a lower or upper bound may be suucient. In this paper we present a datastructure based on ROBDD's, that allows to trade space and time eeciency for precision. The basic extension is provided by the computation of least upper resp. greatest lower bounds of a boolean function under a size constraint for its ROBDD. As a rst practical application we conclude some experimental results for the computation of local don't cares in combinational circuits.