Abstract: The elementary theory of the particle system coincides with the objective laws of the natural world. As a result, the particle system can be used for fireworks and other complex phenomena simulation. To solve the problem that the simulation of particle system is of huge computation and memory resources consumption, the paper built the basic particle system model based on Compute Unified Device Architecture (CUDA) framework. The storage and movement update of particles in the model were considered. Then the parallel KD-TRIE neighbor particle search algorithm based on CUDA was studied. Finally, the detailed implementation of the fireworks simulation was discussed based on the CUDA particle system. The results show that the model can simulate the rise and bloom of the fireworks realistically with the frame rate of up to 320 frames per second, enhancing the fidelity and real-time performance of the simulation.

Key words: Compute Unified Device Architecture (CUDA), particle system, K-Dimensional TRIE (KD-TRIE), fireworks, simulation

摘要： 粒子系统的基本理论符合自然世界的客观规律，适用于烟花等复杂现象的仿真。针对粒子系统仿真存在的计算和内存资源消耗巨大等问题，建立了基于统一计算设备架构(CUDA)框架的粒子系统基本模型，讨论了模型中粒子的存储与运动更新，研究了基于CUDA的并行KD-TRIE邻近粒子搜索算法。运用建立的CUDA粒子系统，研究了烟花仿真的具体实现。仿真结果表明，系统能够较逼真地仿真烟花的上升与绽放过程，帧率高达每秒312帧，仿真的真实感和实时性得到增强。

关键词: 统一计算设备架构, 粒子系统, K维二叉树, 烟花, 仿真