Evolutionary Design of Approximate Sequential Circuits at RTL Using Particle Swarm Optimization

Abstract

Evolutionary circuit design has the ability to explore a wide part of the design space and can lead to satisfactory circuits without human experience and knowledge. In this work, we use Multi-Objective Particle Swarm Optimization to evolve approximate sequential circuits at Register-Transfer Level. A circuit is represented by a two-dimensional array. We aim to produce functional approximate circuits having good trade-off between accuracy, delay and area. The results show the efficiency of the proposed approach.