Abstract:

Using intelligent optimization algorithms as a method to solve the problem of floorplanning in the design of VLSI has been a trend for so many years. In order to solve the slow convergence speed and low efficiency of optimization for the simulated annealing (SA) algorithm and to improve the search strategy and probabilistic inferior transfer, combing the features of the floorplanning problem for heterogeneous 3D NoCs, we adopt the B*tree to indirectly describe the structure of the answer to floorplanning problems. We then introduce the improved SA algorithm into the PSO algorithm, and the resulting hybrid algorithm possesses the advantages of PSO's parallel computing and the SA's global optimization. Simulation results validate the superiority of the new hybrid algorithm to the traditional SA.

Key words: floorplanning；NoC；simulated annealing；particle swarm optimization