Empirical Study of Block Placement by Cluster Refinement

In this paper we propose an efficient cluster refinement approach for macro-cell placement. The algorithm selects a cluster of blocks dynamically, and finds an optimal solution for all the blocks in the cluster simultaneously. This is different from previous zone refinement approach which optimizes the allocation of one single block in each operation. Experimental results on the MCNC benchmark circuits show that the approach achieves excellent area utilization while minimizing the wire length at the same rime.