A new optimization driven clustering algorithm for large circuits

Abstract

It is well known that doing clustering before cell placement could improve the quality of the placement and reduce the run time significantly. The authors present a clustering algorithm which is specially suitable for large designs. It uses local connectivity information to do clustering, and uses global connectivity information to do tie-breaking. Large scale real world circuits show that by this method the improvement could be up to 41% compared to the clustering method without the tie-breaker.<>