Abstract: Software/hardware partitioning is a key problem during the process of system on chip（SOC）design． The partitioning algorithm decomposes the system specification into two sets： one set is marked as candidates for hardware implementation，the other one is marked as software implementation．Many approaches have been presented to tackle this problem，such as mixed integer linear programming（MILP），simulated annealing，tabu search． In this paper，we introduce genetic algorithm to address the problem．Some implementation methods of the algorithm are discussed in detail，including coding，decoding，selection of fitness function，the implementation of selection operator，crossover operator and mutation operator and determination of convergence rule．The comparison with research results in references is given and good results have been gotten by some experiments．