New Methodologies for Congestion Estimation and Reduction

In this paper we summarize different methodologies which have been proposed for congestion estimation and reduction in floorplanning and placement steps for large-scale circuits. Moreover, we formulate the problem of congestion and wire-length minimization in a given initial floorplan to alleviate congestion in the early design cycle. A number of flow-based approaches, to concurrently move congested regions to non-congested regions, have been proposed. These techniques are highly inaccurate due to interference between the block movements. Furthermore, large changes in the floorplan will not preserve region and wirelength constraints. To solve this problem, we propose a novel approach by using a minimum perturbation zero slack assignment for distributing the congestion while preserving the wirelength quality of the design. The proposed technique employs an effective congestion estimator and uses a variation of the ZSA for congestion distribution. Our experimental results show that by applying our method we can reduce congestion by 40% while having less than 11% increase in the wirelength on average.