首页> 外文期刊>Concurrency and computation: practice and experience >Implementingmolecular dynamics simulation on the Sunway TaihuLight system with heterogeneous many-core processors
【24h】

Implementingmolecular dynamics simulation on the Sunway TaihuLight system with heterogeneous many-core processors

机译:使用异构多核处理器在双威TaihuLight系统上实现分子动力学仿真

获取原文
获取原文并翻译 | 示例

摘要

In various research of atom and molecule physical movements, molecular dynamics (MD) simulationrnis a common tool to simulate and investigate the real molecular motion. However, it introducesrna significant penalty in performance, power, electricity, and running time. Consequently,rnonce the simulation size scales up and computing demands keep growing, it comes at substantialrncosts in performance and energy usage. In this paper, an optimized MD implementation onrnthe Sunway TaihuLight supercomputerwith heterogeneousmany-core processors is developed tornaddress the abovementioned issues. The Sunway TaihuLight is a totally independently designedrnand developed Chinese supercomputer with a profusely customized integration approach and arnbrand new many-core processor, the SW26010. The computing power mainly supported by thernhomegrown many-core SW26010 processors differs from other existing heterogeneous supercomputers.rnThe SunwayTaihuLight is a heterogeneous supercomputer that ranks first in theworldrnwith its peak performance over 100 PFLOPS. Firstly,we propose a new algorithm based on clusterrnparticles to fit the special architecture of SW26010. Then, three optimization methods of theMDrnsimulation are implemented step by step: parallelized extensions to SW26010, memory-accessrnoptimizations, and vectorization. After these optimization processes, a 14x speedup is achievedrnon a single computing node and yields significant performance and energy improvements. Superiorly,rnalmost 24,000 computing nodes (6,000,000 cores) are applied in our experiments, whichrngain an almost linear speedup.Besides, the proposed methods also can be adapted and fit to otherrnmolecular dynamics codes, even other similar scientific applications.
机译:在对原子和分子物理运动的各种研究中,分子动力学(MD)仿真器是模拟和研究实际分子运动的常用工具。但是,这会给性能,功率,电力和运行时间带来重大损失。因此,随着仿真规模的扩大和计算需求的不断增长,在性能和能耗上都付出了巨大的代价。针对上述问题,本文在Sunway TaihuLight超级计算机上开发了一种优化的MD实现,该超级计算机具有异构的多核处理器。 Sunway TaihuLight是一款完全独立设计和开发的中国超级计算机,具有大量定制的集成方法和全新的多核处理器SW26010。本地生产的多核SW26010处理器主要支持的计算能力不同于其他现有的异构超级计算机。SunwayTaihuLight是一款异构超级计算机,其最高性能超过100 PFLOPS。首先,我们提出了一种基于簇粒子的新算法,以适应SW26010的特殊架构。然后,逐步实现了三种MD仿真的优化方法:对SW26010的并行扩展,内存访问优化和向量化。经过这些优化过程,单个计算节点的速度提高了14倍,并显着提高了性能和能耗。优越的是,在我们的实验中几乎使用了24,000个计算节点(6,000,000个核),从而获得了几乎线性的加速。此外,所提出的方法还可以适用于其他分子动力学代码,甚至适用于其他类似的科学应用。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号