强激光与粒子束, 2012, 24 (9): 2225, 网络出版: 2012-09-12
适用于Linux系统的高效可并行电磁PIC计算技术
Highly efficient parallel computing technique for electromagnetic PIC simulation in Linux system
摘要
为了能够使大型机和高性能集群得到更好的加速比和并行计算效率,在深入研究并行和FDTD-PIC算法的基础上,在CHIPIC软件平台上开发了Linux系统下的大型全三维可并行电磁PIC代码。分别以一个磁绝缘线振荡器、回旋管和相对论速调管为例进行测试和计算,得到了3.70,6.72和6.00的加速比,验证了在同等的进程数下,高性能集群能够得到比PC机群高得多的加速比和计算效率,能够更好地利用计算资源。
Abstract
For better computing efficiency in high-performance clusters, a three-dimensional electromagnetic PIC parallel code was developed in Linux system based on the CHIPIC software code. By testing a magnetically insulated transmission line oscillator, gyrotron and relativistic klystron, speedups of 3.70, 6.72, 6.00 are obtained for them respectively. Results show that the high-performance clusters have much higher speedups and computing efficiency than the PC clusters with the same processes, so high-performance clusters could make better use of computing resources.
彭凯, 夏蒙重, 刘大刚, 周俊. 适用于Linux系统的高效可并行电磁PIC计算技术[J]. 强激光与粒子束, 2012, 24(9): 2225. Peng kai, Xia Mengzhong, Liu Dagang, Zhou Jun. Highly efficient parallel computing technique for electromagnetic PIC simulation in Linux system[J]. High Power Laser and Particle Beams, 2012, 24(9): 2225.