A hybrid evolutionary analogue module placement algorithm for integrated circuit layout designs

Abstract

This paper presents an integrated approach of simulated annealing (SA) and genetic algorithm (GA) for the analogue module placement in mixed-signal integrated circuit layout designs. The proposed algorithm follows the optimization flow of a normal GA controlled by the methodology of SA. The bit-matrix chromosomal representation is employed to describe the location and the orientation of modules. Compared with the conventional bit-string representation, the proposed chromosomal representation tends to significantly improve the search efficiency. In addition, a slide-based flat scheme is developed to transform an absolute co-ordinate placement of modules to a relative placement. In this way, the symmetry constraints imposed on analogue very large scale integration circuits can be easily fulfilled in the placement run. Use of a radiation-decoder can also drastically shrink the configuration space without degrading search opportunities. The proposed algorithm has been tested with several example circuits. The experiments show this promising algorithm makes the better performance than the simpler SA or GA approaches working alone, and the quality of the automatically generated layouts is comparable to those done manually. Copyright © 2005 John Wiley & Sons, Ltd.