The Research Of Multi-objective PSO For VLSI Floorplan

Integrated circuits play an important role in enhancing the comprehensive nationalstrength. Floorplan is an important issue in the physical design of integrated circuits, asit determines the cost, performance and other properties of chips. It usually aims atmultiple objectives and some of its sub problems have been proved as NP-Complete orNP-Hard. As integrated circuits grow into Very Large Scale Integrated circuits (VLSI)and Ultra Large Scale Integrated circuits (ULSI), the difficulty for floorplan increasesdramatically, and has beyonded the traditional methods. Therefore, it is imperative todevelop the Electronic Design Automation (EDA) technology.Particle Swarm optimization (PSO) is a simple and effective algorithm with agood global searching ability. This thesis focused on the Building Block Layout infloorplan, analyzed the code and multi-objective optimization problems, and developeda multi-objective PSO to give a resolution. The main works are as follow:(1) To deal with floorplan, a suitable representation of placement is necessary.This thesis developed two discrete PSOs with two common representations, andredefined the update formula of particles accordingly. A comparation was donebetween the results of experiments to determine a better one for further study.(2) In view of the deficiency of traditional methods, this thesis introduced theconcept of Pareto and other strategies into PSO, then proposed a multi-objective PSOfor floorplan, which minimized the layout area and total interconnection wire lengthsimultaneously.(3) Coevolutionary algorithm is superior to evolutionary algorithm. Havingstudied of the coevolution and multi-objective coevolutionary algorithms, this thesisproposed a coevolutionary multi-objective PSO based on the existing algorithm. Theexperiment results showed the new algorithm performed better.