A high-level test synthesis based on genetic algorithm is proposed. This can give consideration to the testability problem in process of scheduling, module allocating and register allocating. Meanwhile, a chromosome coding of genetic algorithm is proposed, it can be used for high-level scheduling and module allocating simultaneously. A one point crossover operator based on data dependence and a mutation operator based on control step constrain are designed avoiding generation of infeasible solutions. The efficiency of testability improvement has been demonstrated by experiment.