并行多块结构重叠网格装配算法及应用
针对多块结构重叠网格并行装配的问题,设计了支持初始网格系统细分的多块结构重叠网格框架,并在此框架基础上提出了基于局部洞映射的并行挖洞算法、格心网格下可跨块寻点的并行搜索算法,使之可适应大规模并行数值模拟时的分布式计算环境。此算法被模块化地集成到了自主研发的大规模多块结构网格数值求解器(CCFD-MGMB)中,可支持大规模并行非定常多体分离数值模拟。并行测试结果表明,算法具有良好的局部数据结构组织,数据可扩展性强。数值应用模拟结果表明了该算法的有效性及正确性,千核并行非定常数值计算效率(相对于64核)可达58%。...
Saved in:
| Published in | 计算机应用研究 Vol. 33; no. 3; pp. 788 - 793 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
中国科学院计算机网络信息中心 超级计算中心,北京 100190
2016
数学工程与先进计算国家重点实验室,江苏 无锡 214125 中国科学院大学,北京 100190%中国科学院计算机网络信息中心 超级计算中心,北京,100190%中国科学院计算机网络信息中心 超级计算中心,北京 100190 |
| Subjects | |
| Online Access | Get full text |
| ISSN | 1001-3695 |
| DOI | 10.3969/j.issn.1001-3695.2016.03.033 |
Cover
| Summary: | 针对多块结构重叠网格并行装配的问题,设计了支持初始网格系统细分的多块结构重叠网格框架,并在此框架基础上提出了基于局部洞映射的并行挖洞算法、格心网格下可跨块寻点的并行搜索算法,使之可适应大规模并行数值模拟时的分布式计算环境。此算法被模块化地集成到了自主研发的大规模多块结构网格数值求解器(CCFD-MGMB)中,可支持大规模并行非定常多体分离数值模拟。并行测试结果表明,算法具有良好的局部数据结构组织,数据可扩展性强。数值应用模拟结果表明了该算法的有效性及正确性,千核并行非定常数值计算效率(相对于64核)可达58%。 |
|---|---|
| Bibliography: | 51-1196/TP Ma Wenpeng,Lu Zhonghua,Yuan Wu,Liang Shan(1. Supercomputing Center, Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China; 2. University of Chinese Academy of Sciences, Beijing 100190, China; 3. State Key Laboratory of Mathematical Engineering & Advanced Computing, Wuxi Jiangsu 214125, China) This paper proposed a framework of multi-block overset grid that supports block-splitting from the initial grid system for the problem of parallel multi-block overset grid assembly. And based on this framework,the present methods developed a parallel local hole-mapping approach and a cell-centered based donor searching algorithm. The developed methods supported inter-block donor searching and were suitable for the distributed environment where the large-scale numerical simulations were launched. These methods were integrated as a module to the domestically developed multi-block solver called CCFD-MGMB and made the solver capable of simulating multi-body separation unsteady probl |
| ISSN: | 1001-3695 |
| DOI: | 10.3969/j.issn.1001-3695.2016.03.033 |