Two Clustering Preprocessing Techniques for Large-Scale Circuits

In this paper, two effective preprocessing techniques for clustering large-scale circuits are presented. These techniques can be performed before the general multilevel clustering to enhance the speed of the partitioning technique while preserving the solution quality. A "wrapped" version of hMETIS is implemented, in which the proposed techniques are applied as the preprocessing step. The empirical results on standard benchmark circuits show that the application of the proposed techniques improves the over all runtime by 30% and the partitioning results by 2%.