Flow-based cell moving algorithm for desired cell distribution

We propose a new cell moving algorithm to achieve a desired cell density distribution during the detailed placement stage of a large size standard-cell placement flow. To evaluate our algorithm, we also define a distribution deviation metric on how close a placement result is to the target cell density distribution. Experiments show that our approach can generate placement results, which have a cell density distribution closer to the desired distribution with very little increase in the wirelength. We achieve a 32.3% improvement in the distribution deviation at the cost of 11.2% increase in the wirelength on average. Also, the final placement results have 5.9% shorter wirelength than those of an industrial placer, Cadence QPlace.