【作者】 常新正； 王宇新； 狄少丞； 郭禾； 季顺迎；

【Author】 Chang Xin-Zheng 1 Wang Yu-Xin 1 Di Shao-Cheng 2 Guo He 1 Ji Shun-Ying 2 1 School of Computer Science and Technology, Dalian University of Technology, Dalian 116023, China 2 State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian 116023, China

【机构】 大连理工大学计算机科学与技术学院； 大连理工大学工业装备结构分析国家重点实验室；

【摘要】 大规模颗粒离散元计算在研究地质灾害的形成和发生时有重要的意义。本文基于颗粒离散元基本原理,借助CUDA编程平台实现了基于GPU的大规模颗粒离散元并行计算,包括颗粒与颗粒、颗粒与墙碰撞模型计算方法的并行化,计算的颗粒数目达到100万以上。系统通过优化并行计算模型和内存访问模式提高了计算速度和效率,在颗粒数目为5万时获得了最高36倍的计算加速比,并且加速比随颗粒数目的增加呈线性增加,进一步验证了GPU离散元计算的有效性、高效性和稳定性。本文的研究为下一步颗粒离散元大规模计算的工程应用提供了基础。更多还原

【Abstract】 Particle discrete element simulation has important significance for the study of formation and occurrence of geological disasters. Based on the basic principles of particle discrete element, a GPU-based parallel computation method of large-scale particle dynamics simulation is presented in this paper. With the CUDA programming platform, both collision models among particles and between particles and the walls are calculated. When the particles size up to 50,000, a 36 times speedup is achieved. The scale of the particles can reach more than 1 million and with the particles increase, the speedup number tends to linear growth. This method verifies the effectiveness, efficiency and stability of the GPU-based particle discrete element computation and provides a basis for its engineering application.更多还原