This paper presents a new algorithm, hybrid algorithm (HA), to improve the performance in circuit partitioning while using less resources. Although there have already been two circuit partitioning techniques widely used, SA and GA, each of them have advantages and disadvantages. On the one hand, SA does not requires much memory but takes a lot of time to perform. On the other hand, GA performs faster but needs more memory. The proposed algorithm combines both of them together and produces a superior result. It improves the speed while using moderate storage. HA starts from finding an initial solution from a relatively good candidate solution. Then a new solution is created through mutation and the cost function of the new solution is compared with that of the initial solution. After the algorithm makes a decision to keep the old solution or the new one, a newer solution is created again and the cost function is compared with that of the formerly selected solution. This process will iterates until the temperature becomes zero. Experiments were done on a PC using hypergraphs from ISPD98 Circuit Benchmark Suite. SA, GA and our proposed algorithm are run on the benchmark to compare average cost function, minimum cost function and average CPU time. The results show the superior performance of HA compared to the traditional optimization techniques.
[1]
Byung Ro Moon,et al.
Genetic Algorithm and Graph Partitioning
,
1996,
IEEE Trans. Computers.
[2]
Naveed A. Sherwani.
VLSI Physical Design Automation
,
1995
.
[3]
Chung-Kuan Cheng,et al.
Ratio cut partitioning for hierarchical designs
,
1991,
IEEE Trans. Comput. Aided Des. Integr. Circuits Syst..
[4]
R. J. Mack,et al.
VLSI physical design automation: theory and practice
,
1996
.
[5]
Theodore W. Manikas,et al.
Genetic Algorithms vs. Simulated Annealing: A Comparison of Approaches for Solving the Circuit Partitioning Problem
,
1996
.