Buffer and wire-size optimization under higher order RLC model for interconnect design

In this paper, we study the interconnect optimization problem under a higher-order distributed RLC model to optimize not just area and delay, but also waveform and crosstalk for RLC circuit with non-monotone signal response. We propose a new multi-objective genetic algorithm (MOGA) which uses a single objective sorting (SOS) method for constructing the non-dominated set to solve this multi-objective interconnect optimization problem. The MOGA/SOS optimal algorithm provides a smooth trade-off among signal delay, wave form, and routing area. Extensive experimental results show that our algorithm is scalable with problem size. Furthermore, compared to the solution based on an Elmore delay model, our solution reduces the total routing area by up to 30%, the delay to the critical sinks by up to 25%, while further improving crosstalk up to 25.73% on average.