The greedy genetic algorithm is proposed to solve the hybrid flow shop scheduling problem with unrelated parallel machines for minimizing makespan. Firstly, the chromosome coding is based on the job processing sequence and two kinds of machine allocation schemes are considered. Considering the influence of different stages’ equipment configuration on the result of scheduling, the forward or reverse decoding strategy with the corresponding rescheduling method are used. Moreover, the greedy crossover and mutation operators are proposed. They not only improve the quality and diversity of the population, but also have a strong local search ability. Finally, the orthogonal experiment is done to determine the parameters of the algorithm, the computation results on the known cases show that the effectiveness of the algorithm compared with the known algorithms. At the same time, the experiments also show the necessity of the forward and reverse decoding strategy and the timing of using them.