| 基于CUDA的离散粒子系统模拟仿真及其实现 |
| |
| 引用本文： | 江顺亮,黄强强,董添文,徐少平.基于CUDA的离散粒子系统模拟仿真及其实现[J].南昌大学学报(工科版),2011,33(3):290-294. |
| |
| 作者姓名： | 江顺亮 黄强强 董添文 徐少平 |
| |
| 作者单位： | 1. 南昌大学信息工程学院,江西南昌,330031
2. 上饶师范学院物理与电子信息系,江西上饶,334001 |
| |
| 基金项目： | 国家自然科学基金资助项目(50863003) |
| |
| 摘 要： | 使用CUDA（compute unified device architecture，统一计算架构），利用GPU（graphic processing unit，图像处理器）强大的并行能力实现DEM（discrete element method，离散单元法）的离散粒子系统模拟仿真。邻域搜索时使用基于哈希表排序的邻域搜索算法，依据CUDA内存模型的特点对系统做出有效的内存规划，并使用CUDA流技术通过将粒子碰撞和数据输出分割成4个流异步执行，来隐藏数据拷贝的部分延迟。通过使用CUDA的事件作为GPU计时器监测，实验证明本文中的并行搜索算法的执行效率很高，而使用合理的内存策略和流技术也可以有效的提高系统盼性能。
|
| 关 键 词： | GPU 并行程序设计 CUDA 粒子系统 流 |
Simulation and Implementation of Discrete Particle System based on CUDA |
| |
| JIANG Shun-liang,HUANG Qiang-qiang,DONG Tian-wen,XU Shao-ping.Simulation and Implementation of Discrete Particle System based on CUDA[J].Journal of Nanchang University(Engineering & Technology Edition),2011,33(3):290-294. |
| |
| Authors: | JIANG Shun-liang HUANG Qiang-qiang DONG Tian-wen XU Shao-ping |
| |
| Affiliation: | JIANG Shun-liang1,HUANG Qiang-qiang1,DONG Tian-wen2,XU Shao-ping1(1.School of Information Engineer,Nanchang University,Nanchang 330031,China,(2.Department of Physics,Shangrao Normal School,Shangrao 334001,China) |
| |
| Abstract: | In this paper,NVIDIA's parallel computing architecture(CUDA) is used to implement a discrete particle simulation system.In the neighbor search part,an algorithm is used based on hash list.In the collision part,the stream of CUDA is used to split one step simulation to four parallel streams;the asynchronous copy method is used to output;then an efficient way to use the CUDA memory model is designed.The results showed that the CUDA's discrete particle system is efficient and effective. |
| |
| Keywords: | GPU parallel programing CUDA particle system stream |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
