New metaheuristic algorithms for non-slicing VLSI floorplanning

Abstract

A classical problem in VLSI floorplanning is to arrange a set of rectangular modules on a rectangular chip area ensuring that no two modules overlap with each other which provide an optimum measure of performance. In our thesis we incorporate different metaheuristic algorithms to solve the VLSI floorplanning problems which produce different results according to the applied algorithms. While applying these algorithms we incorporate the use of an effective method that is Clonal Selection Algorithm (CSA) with mutation operator to explore the search space. Then these algorithms also use a number of single state methods separately, to exploit the search space. This incorporation of CSA instead of other methods like crossover is a novel work in this area. Experimental results show that our metaheuristic algorithms provide better result than those algorithms which use other metaheuristic algorithms. In our thesis, we use the existing popular benchmarks with a view to comparing our results with the previous research works. Hence, our algorithms produce optimal or nearly optimal solutions for all the tested benchmark problems. Moreover, we develop visualization software, incorporating all our algorithms, along with graphical output of floorplans, which can show different comparative results on basis of statistical test and Kruskal-Wallis test for different applied algorithms.