| HL-2A��������ʵʱƽ���ؽ���GPU���л� |
| |
| 引用本文： | 马瑞,夏凡,凌飞,李佳鲜. HL-2A��������ʵʱƽ���ؽ���GPU���л�[J]. 核聚变与等离子体物理, 2017, 37(2): 159-166. DOI: 10.16568/j.0254-6086.201702006 |
| |
| 作者姓名： | 马瑞 夏凡 凌飞 李佳鲜 |
| |
| 作者单位： | (?????????????о??????? 600041) |
| |
| 基金项目： | 国家磁约束核聚变能发展研究专项 |
| |
| 摘 要： | 描述了HL-2A等离子体实时平衡重建的GPU并行化算法,主要包括G-S方程的并行化处理、三对角方程求解、网格边界磁通计算以及一系列矩阵相乘的并行加速.并行后,在129×129的网格下完成一次迭代计算需要约575μs.
|
| 关 键 词： | 关键字:HL-2A装置 平衡重建 图形处理器 通用计算架构 |
GPU parallelization of HL-2A plasma real-time equilibrium reconstruction |
| |
| MA Rui,XIA Fan,LING Fei,LI Jia-xian. GPU parallelization of HL-2A plasma real-time equilibrium reconstruction[J]. Nuclear Fusion and Plasma Physics, 2017, 37(2): 159-166. DOI: 10.16568/j.0254-6086.201702006 |
| |
| Authors: | MA Rui XIA Fan LING Fei LI Jia-xian |
| |
| Affiliation: | (Southwestern Institute of Physics, Chengdu 610041) |
| |
| Abstract: | The GPU parallel algorithms of the HL-2A plasma real-time equilibrium reconstruction including parallel processing of G-S equation, an algorithm solving the tri-diagonal linear system efficiently in parallel, the computation of the flux on grid boundary points and some parallel speed-up for middle-scale matrix multiplication on GPU, were intoduced. After the parallelization, an equilibrium reconstruction iteration needs around 575μswith a 129×129 grid. |
| |
| Keywords: | HL-2Atokamak Equilibrium reconstruction GPU CUDA |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《核聚变与等离子体物理》浏览原始摘要信息 |
|
点击此处可从《核聚变与等离子体物理》下载全文 |
