Optimization of Hybrid and Local Search Algorithms for Standard Cell Placement in VLSI Design

Optimization of Non-Deterministic Polynomial hard (NP-hard) problems of non-trivial sizes is done using heuristic approach. Until now, simulated annealing (SA), genetic algorithm (GA) and Hopfield neural network (HNN) were individually used for solving the standard cell placement (SCP) problem. Now, in this paper we discuss systems based on hybridizing HNN, SA and GA. We begin with HNN & GA hybrid system. Here, truncated HNN places 90% cells in 10% computation time followed rest by GA without disturbing the position of already placed cells. We then extend to SA & GA hybrid systems that use either a coupling mechanism or integrating the methods with some features of them to evolve a new method, Parallel Recombinative SA (PRSA) wherein a solution representation is chosen followed by development of cross-over and neighborhood operators. Finally parallelization of PRSA is proposed and taken up. Here the problem is divided into sub-problems and assigned to different processors. For the best computational efficiency, these sub-problems should be least interdependent. In the end we compare them in terms of solution quality and computing speed in connection with the standard cell placement problem followed by comparing the results of the proposed technique with the established results of local search ones such as Tabu search and Threshold accepting.