Smooth particle hydrodynamics (SPH) has a unique advantage in the study of wind and sand due to its meshlessness. When the SPH method is used to study the wind and sand flow, it is necessary to discretize the entire calculation area into a large number of single particles, so the calculation scale is large and the calculation efficiency is low. In order to improve the computational efficiency of SPH method, CUDA platform which supports parallel computing and GPU massively parallel computing technology were adopted to realize the accelerated operation of SPH method in numerical simulation. Taking the twodimensional airsand twophase coupling model as a numerical example, the particle swarm movement law was analyzed in detail by using GPU parallel computing. The computational efficiency of CPU and GPU under different particle numbers and the influence of the number of GPU threads on the computational efficiency were compared. After statistical analysis of the results, the typical parabolic shape and the variant saltation trajectories were obtained. The simulation results show that the SPH-GPU parallel computing technology can be applied to the numerical simulation research of the wind and sand structure.