Performance study of an implicit Navier-Stokes solver on a many-core machine
文献类型：会议论文
| 作者 | Zhengzheng Yan; Rongliang Chen; Li Luo; Ziju Liao; Xiao-Chuan Cai |
| 出版日期 | 2017 |
| 会议日期 | 2017 |
| 会议地点 | 亚特兰大 |
| 英文摘要 | In this talk, we present our resent works on the implicit Navier-Stokes solver on the newest many-core machine Sunway TaihuLight supercomputer, which uses many-core processors that consist of management processing cores (MPCs) and clusters of computing processing cores (CPCs). In our approach, we solve the 3D time-dependent Navier-Stokes equations on unstructured tetrahedral meshes using a fully implicit Newton-Krylov-Schwarz (NKS) method. The implicit Navier-Stokes solvers on unstructured meshes are challenge due to the challeng- ing of solving the large scale ill-conditioned linear/non-linear systems at each time-step, indirect addressing, large bandwidth, non-unit stride memory access patterns, and so on. Thus, it is interesting to study the execution amenability of them on the state-of-the-art many-core architecture. In this talk, we will focus on the re-factor and optimization techniques of the shared memory paral- lelization of the unstructured nite element method and the subdomain solver of the NKS method on the many-core processors, including the overlapping of the computing and data moving on the CPCs, the overlapping of the MPCs and CPCs computing, the minimizing of the data moving, fully utilizing of the lim- ited on-chip fast bu er, and so on. Finally, as an application example, we test the algorithm for a blood ow simulation with realistic geometries and realistic parameters with over 16k MPCs and one million CPCs. |
| 语种 | 英语 |
| 源URL | [http://ir.siat.ac.cn:8080/handle/172644/12691]  |
| 专题 | 深圳先进技术研究院_数字所 |
| 作者单位 | 2017 |
| 推荐引用方式 GB/T 7714 | Zhengzheng Yan,Rongliang Chen,Li Luo,et al. Performance study of an implicit Navier-Stokes solver on a many-core machine[C]. 见:. 亚特兰大. 2017. |
入库方式： OAI收割
来源：深圳先进技术研究院
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。