摘要：To solve the hardware/software(HW/SW) partitioning problem on the system that contains only one CPU, a new algorithm based on GA is studied. Firstly, the concept of hardware orientation is put forward, and then used to create the initial colony of GA and in mutation process, which reduces the randomicity of initial colony and the blindness of search. Secondly in the process of GA, crossover and mutation probability become more and more small, this not only ensures a big search space in the early, but also keeps the good solution later. Experimental statistics show that the efficiency of the proposed algorithm outperforms the algorithms in comparison by up to 23% in large-scale problem. What’s more, it can obtain better solution. In conclusion, the proposed algorithm has higher efficiency and appears to be a better solution under specific conditions.
