节点文献

基于天河二号超算的网格无关性及并行研究

Grid independence and parallel study based on Tianhe Ⅱ supercomputing

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 徐艺峰李健王杰范冰丰王钢

【Author】 XU Yi-feng;LI Jian;WANG Jie;FAN Bing-feng;WANG Gang;School of Electronics and Information Engineering,Sun Yat-sen University;Institute of Advanced Technology,Sun Yat-sen University;State Key Laboratory of Optoelectronic Materials and Technologies,Sun Yat-sen University;

【机构】 中山大学电子与信息工程学院中山大学先进技术研究院中山大学光电材料与技术国家重点实验室

【摘要】 以CFD商用软件Fluent 15.0为基础,天河二号超级计算机为平台,ZnO-MOCVD腔体为数值模型,对网格无关性及超算中心的并行计算能力进行研究。通过计算7种不同网格数量的ZnO-MOCVD模型的沉积率,确定网格数量为408万时沉积率结果达到网格无关性。根据工程需要选取其中两种模型,设置并分析多种并行组合,得出网格数量为17万和408万的MOCVD腔体模型的最佳处理器核数分别为16和128,最大加速比分别可以达到9和45,验证了该软件在天河二号上并行处理的实用性,结果可为后续大规模计算提供参考。

【Abstract】 Based on the Fluent 15.0,the platform of Tianhe II supercomputer,and the numerical model of ZnO-MOCVD chamber,the grid independence and the parallel computing capabilities of the supercomputing center were studied.By calculating the deposition rate of ZnO-MOCVD model with seven different grid numbers,the number of grids whose deposition rate resulting in grid independence was determined to be 4 million 80 thousand.According to the needs of the project,two models were selected,and a number of parallel combinations was set up and analyzed,and the optimal processor cores of the MOCVD chamber model with 170 thousand mesh number and 4 million 80 thousand mesh number were obtained,respectively,16 and 128,and the maximum speedup reached 9 and 45 respectively.By comparing the computational efficiency,the optimal parallel computing combination was obtained,and the simulation cycle was shortened greatly,the practicality of Fluent parallel processing on Tianhe II was verified,the results provided a reference for following large-scale computation.

  • 【文献出处】 计算机工程与设计 ,Computer Engineering and Design , 编辑部邮箱 ,2018年07期
  • 【分类号】TP338
  • 【被引频次】7
  • 【下载频次】279
节点文献中: 

本文链接的文献网络图示:

本文的引文网络