Multiobjective evolution of approximate multiple constant multipliers

Multiple constant multiplier (MCM) is a digital circuit which multiplies its single input by N constants. As MCMs are composed of adders and shifters, their implementation cost is relatively low. In this paper, we propose a method for design of approximate multiple constant multipliers where the requirement on functional equivalence between the specification and implementation is relaxed in order to further reduce the area on a chip or minimize delay. The proposed method is based on multiobjective Cartesian Genetic Programming. It provides many trade-off solutions among accuracy, area and delay.