Towards optimizing global mincut partitioning

Abstract

Mincut algorithms have received much attention in the past for treating the placement problem in layout synthesis. The paper introduces a new class of mincut partitioning algorithms (SQP) meeting global minimization requirements. The new class of algorithms in its different variations is empirically compared with the classical mincut procedures as well as with recent extensions. The new algorithms have shown a significant (10 to 40%) improvement in the overall netlength compared with known algorithms. Moreover, the new class of algorithms is proved to have a linear time complexity.<>