HARDWARE/SOFTWARE PARTITIONING ALGORITHM BASED ON THE COMBINATION OF GENETIC ALGORITHM AND TABU SEARCH

To solve the hardware/software (HW/SW) partitioning problem of a single Central Processing Unit (CPU) system, a hybrid algorithm of Genetic Algorithm (GA) and Tabu Search(TS) is studied. Firstly, the concept hardware orientation is proposed and then used in creating the initial colony of GA and the mutation, which reduces the randomicity of initial colony and the blindness of search. Secondly, GA is run, the crossover and mutation probability become smaller in the process of GA, thus they not only ensure a big search space in the early stages, but also save the good solution for later browsing. Finally, the result of GA is used as initial solution of TS, and tabu length adaptive method is put forward in the process of TS, which can improve the convergence speed. From experimental statistics, the efficiency of proposed algorithm outperforms comparison algorithm by up to 25% in a large-scale problem, what is more, it can obtain a better solution. In conclusion, under specific conditions, the proposed algorithm has higher efficiency and can get better solutions.