The Unified Gas-Kinetic Wave-Particle (UGKWP) method combines the advantages of deterministic and particle-based approaches through a wave-particle framework, enabling efficient simulation of flows across all flow regimes and speed ranges, demonstrating its potential as a tool for multiscale flow simulations. However, com-pared to traditional Computational Fluid Dynamics (CFD) methods, the UGKWP method requires more accurate geometric information and more complex data structures, necessitating preprocessing of input data files prior to computation. As the grid size increases, the limited parallelism of the UGKWP preprocessing method leads to a significant decline in computational efficiency. To address this issue, this study proposes a parallel preprocessing method for two-dimensional unstructured UGKWP based on a GPU heterogeneous platform using CUDA programming. The algorithm migrates both serial and computational components to the highly parallel GPU platform and employs atomic operations to resolve data conflicts, thereby improving computational accuracy and achieving parallelization of the preprocessing method. Numerical experiments demonstrate that the parallel preprocessing method achieves a speedup of 10–12 times, exhibiting excellent parallel performance.