Buffer planning algorithm based on partial clustered floorplanning

Abstract

In this paper, we propose a partial clustered floorplanning methodology with buffer planning. The theoretic analyses show that the timing constraints can be transferred into bounding box constraint and the spacing between buffers is somewhat stable. Therefore the critical nets can be controlled by the clustering strategy. The cluster strategies in our approach are designed not only for localizing the critical nets, but also for facilitating the buffer insertion of long wires. Based on the CBL representation, we devise sub CBL to represent the cluster and embed the optimization of the clusters into the annealing process. In most of the previous clustering-based methods, the shape of the cluster was restricted to a square. In this paper, however, we remove this restriction by treating the cluster as the sub packing. Our method can achieve a very stable performance. Experimental results on the MCNC benchmark show the effectiveness of the method and prove the correctness of the theoretic analyses.