摘 要:针对目前 WRF 模式迁移中存在的问题,提出具体迁移部署的方法。随后通过对美国东部地区降雨模拟,给出WRF 模式在 Intel X86 和 ARM 架构上的计算性能的差异,通过 NCL 方法计算了结果的准确性。实验结果表明,WRF 模式可被完整地迁移到 ARM 架构的超级计算机上。仿真结果显示,在并行运算下,WRF 模式在 ARM 架构超算集群上耗时更短,且单节点运算中 ARM 可有效提高并行运算效率,具有一定的使用价值,为研究者提供参考。
关键词:X86 架构;ARM 架构;鲲鹏 920 处理器;WRF 模式;移植优化
DOI:10.19850/j.cnki.2096-4706.2023.06.020
中图分类号:TP391.9 文献标识码:A 文章编号:2096-4706(2023)06-0077-05
Research on Migration and Optimization of WRF on ARM Server
CAO Renzhi, WEI Tianxiang
(CHN Energy Digital Intelligence Technology Development (Beijing) Co., Ltd., Beijing 100011)
Abstract: Aiming at the problems existing in WRF migration, a concrete method of migration and deployment is proposed. Subsequently, by simulating the rainfall in the eastern region of the United States, the difference of computing performance of WRF on Intel X86 and ARM architecture is given, and the accuracy of the results is also calculated by the NCL method. The experimental results show that WRF can be completely migrated to supercomputer on ARM architecture. The simulation results show that WRF takes less time in the supercomputer cluster on ARM architecture under parallel operation, and ARM can effectively improve the efficiency of parallel operation in the single-node operation, which has certain application value and provides a reference for researchers.
Keywords: X86 architecture; ARM architecture; Kunpeng 920 processor; WRF; migration and optimization
参考文献:
[1] GEIST A, REED D A. A survey of high-performance computing scaling challenges [J].Int J High Perform Comput Appl,2017,31(1):104-113.
[2] GWENNAP L. ThunderX2 Strengthens ARM Servers: Cavium Starts Production of World's Most Powerful ARM CPU [J]. Microprocessor report, 2018, 32(5):1,4-6.
[3] 新华社 . 日本富士通发布新型超算“大脑”[J]. 中国建设信息化,2018(16):7.
[4] PADOIN E L,OLIVEIRA D,VELHO P,et al. Evaluating Performance and Energy on ARM-based Clusters for High Performance Computing [C]//2012 41st International Conference on Parallel Processing Workshops.Pittsburgh:IEEE,2012:165-172.
[5] PEDRETTI K,YOUNGE A J,HAMMOND S D,et al. Chronicles of Astra: Challenges and Lessons from the First Petascale Arm Supercomputer [C]//SC20: International Conference for High Performance Computing,Networking,Storage and Analysis.Atlanta,IEEE:1-14.
[6] 莫秋燕,吴家隐,李先绪,等 . 国产处理器及其在服务器中的应用 [J]. 计算机产品与流通,2019(7):176
.[7] TOP500.Japan Captures TOP500 Crown with Arm-Powered Supercomputer None [EB/OL].(2022-06-22).https://www.top500.org/ news/japan-captures-top500-crown-arm-powered-supercomputer.
[8] 胡伯彦,汤剑平,王淑瑜 . 东亚地区晴空湍流未来变化趋势预估:基于 CORDEX-WRF 模式降尺度 [J]. 地球物理学报,2022,65(7):2432-2447.
[9] 张建彬,高志球,杨军,等 . 基于 WRF 模式的博斯腾湖地区暴雨数值模拟研究 [J]. 高原气象,2022,41(4):887-895.
[10] 过怡 , 刘文芝 . 汇编语言差异比较——从 X86 到 ARM[J]. 江苏科技信息,2019,36(22):39-42.
[11] 孙启龙,赵学良,陈芋文,等 . 一种提高 WRF 并行计算效率的方法:CN105938427A[P].2016-09-14.
作者简介:曹人之(1997—),女,汉族,四川绵阳人,网络安全工程师,硕士研究生,研究方向:信息安全,高性能计算等;魏天祥(1998—),男,汉族,北京人,网络安全工程师,硕士研究生,研究方向:信息系统安全,密码学等。