**[Application Background] **For aerodynamic performance simulation and aerodynamic shape optimization of high-speed trains in complex scenes, the goal of our work is to improve the simulation accuracy and parallel computing efficiency. **[Methods]** A high-performance computational fluid dynamics software CCFD v3.0 is independently developed, and the core functional modules such as high-precision NS equation space-discrete algorithm, efficient linear algebra solution, turbulence simulation, and parallel grid processing are developed. It has been ported to two domestic heterogeneous supercomputing platforms and specifically optimized for parallel computing. **[Results] **The computational results delivered by CCFD v3.0 show that the head shape of the high-speed trains, the bottom resistance and friction are the main sources of resistance of the train in running, accounting for 71% and 23%, respectively. The 89% of the resistance of the trains carriages comes from friction and 11% from form drag. **[Conclusions] **CCFD v3.0 achieves good parallel acceleration in prediction of the aerodynamic performance of the high-speed trains and analysis of the flow field mechanism.