Improving Standard Cell Placement through Adaptive Parameter Control

In this work, we present a profile based meta-level framework to adapt the control parameters of stochastic algorithms, like simulated annealing (SA) or genetic algorithm (GA). The framework uses past data about the algorithm's performance and automatically conditions the parameter selection strategies depending on the resource availability. We use the control framework for solving the standard cell placement problem under time constraints. The experimental results show that the placement qualities can be significantly improved using profile based adaptive control.