ParSA: Parallel simulated annealing placement algorithm for multi-core systems

Simulated annealing is known as a widely used algorithm for complex and nonlinear combinatorial optimization problems. This technique has been applied to variety of the problems such as VLSI cell placement. However, simulated annealing placement requires long execution time to brows the search space and find the near-optimal solution. There were many attempts to parallelize it on multi-processor systems but addressed algorithms results in poor speedup due to considerable communication overhead between the processors. In this paper, we proposed a new parallel Simulated Annealing based on multi-core systems with very low communication overhead. Experimental results show the speedup of the proposed algorithm is improved by 32% on average without considerable quality degradation.