Optimization Algorithm for the Unstructured UGKWP Particle Tracking Process Based on a GPU

The Unified Gas–Kinetic Wave–Particle (UGKWP) method is a multiscale method that offers high computational efficiency when solving complex high-Mach-number flows around spacecraft. When the UGKWP method, based on a Graphics Processing Unit (GPU) platform, is used to simulate flow, threads within the same warp are responsible for tracking different particles, leading to a significant warp divergence problem that affects overall computational efficiency. Therefore, this study introduces a dynamic marking tracking algorithm based on block sharing to enhance the efficiency of particle tracking. This algorithm rebuilds the original tracking process by marking and tracking particles, aligning thread computations within the same warp as much as possible to reduce warp divergence. As a result, the average number of active threads increased by over 46% across different testing platforms. The optimized UGKWP platform was used to simulate the re-entry capsule case, and the results showed that the optimized UGKWP can accurately and efficiently simulate the flow details around the capsule. This research provides an efficient and accurate tool for simulating complex multiscale flows at high Mach numbers, which is of great significance.