Efficient Hybrid Search Techniques For Circuit Partitioning

– Recently, four search approaches have emerged for handling complex combinatorial optimization problems: Simulated Annealing, Genetic Algorithms, Tabu Search and Greedy Randomized Search Procedure (GRASP). Based on simple ideas, these heuristics have been extremely efficient in finding near-optimal or optimal solutions for many types of difficult problems [1]. In this paper we apply these approaches to the circuit partitioningproblem highlighting their strengths and weaknesses. Two hybrid search techniques for the partitioningproblem are developed by using GRASP and a Genetic Algorithm to generate good initial partitions. Tabu Search is used to refine the solution quality and reduce the overall computational time of the simpler standalone search techniques. Very promising results show the advantages of combining the strengths of these newer search methods.