Learning as applied to stochastic optimization for standard cell placement

Although becoming increasingly important, stochastic algorithms are often slow since a large number of random design perturbations are required to achieve an acceptable result-they have no built-in “intelligence”. In this work, we used regression to learn the swap evaluation function while simulated annealing is applied to 2D standard-cell placement problem. The learned evaluation function is then applied to the trained simulated annealing algorithm (TSA). The annealing quality improvement of TSA was 15%~43% for the set of examples used in learning and 7%~21% for new examples. With the same amount of CPU time, TSA could improve the annealing quality by up to 28% for some benchmark circuits we tested. In addition the use of the evaluation function successfully predicted the effect of the windowed sampling technique and derived the informally accepted advantages of windowing from the test set automatically