Performance improvement of a genetic algorithm for floorplanning with parallel computing technology

Abstract

This paper presents a parallel implementation of a genetic algorithm for floorplanning. It discusses the strategy and implementation and proposes a Multi-Selection-Multi-Evolution (MSME) parallelization scheme. The initial experimental results on a 12-node IBM SP2 scalable parallel computer have shown that the scheme is effective in improving performance of floorplanning over that of a serial implementation. The parallel version could obtain better results with more than 80% of probability and the improvements on chip area and maximum path delay could be more than 20% in the best cases.