A hierarchical standard cell placement method based on a new cluster placement model

In this paper, we propose a new hierarchical timing-driven standard-cell placement method, in which a new cluster placement model is introduced. The proposed method is divided into three phases: clustering cells, global placement and detailed placement. After clustering cells, in the global placement phase, we determine and improve cluster placement by a simulated annealing based method. In most previous methods, the shape of any cluster was restricted to a square. In the proposed method, however, we remove this restriction, and propose a new placement model, called the Amoeba model, in which the shape of a cluster can be a collection of connected squares. Since the flexibility of cluster placement is increased by the Amoeba model, we can obtain a high quality placement satisfying timing constraints. In the detailed placement phase, we assign cells to cell rows under the non-overlapping constraint of cells with a constructive approach. Experimental results show the effectiveness of the proposed method with the new cluster placement model.