首页> 外文学位 >MDoC: A molecular dynamics application framework for the QCDOC supercomputer.
【24h】

MDoC: A molecular dynamics application framework for the QCDOC supercomputer.

机译:MDoC:QCDOC超级计算机的分子动力学应用框架。

获取原文
获取原文并翻译 | 示例

摘要

We present the first molecular dynamics application package, called MDoC, for the QCDOC supercomputer designed, superficially, for QCD (quantum chromodynamics) applications. Due to the high complexity and uniqueness of the hardware architecture particularly the network, other molecular simulation packages will require major development to take advantage of QCDOC features. MDoC is custom designed, to fully exploit such unique architectural advances.; Typical bio-molecular systems have thousands, possibly millions, of atoms with timescales at microseconds to milliseconds. Simulating them requires a substantial amount of computational power that can be provided by using supercomputers. The most time consuming calculation for simulating bio-molecules is that of the electrostatic interactions, commonly handled through the use of the Ewald algorithm. We also address other algorithms necessary for molecular simulations, specifically, constant temperature and pressure algorithms using the Melchionna-modified Nose-Hoover method. We discuss an efficient all-gather method for QCDOC. Since QCDOC does not have a network specific algorithm for global communication like the tree network on BlueGene/L, we present one specifically for handling the long range electrostatic interactions. We demonstrate the usability of QCDOC to for non-QCD applications. Due to the fast communication network and the flexibility of the PowerPC processor, molecular dynamics simulations are capable of 70% parallel efficiency when using 1024 processors for a protein with approximately 30,000 atoms including solvents.
机译:我们为QCDOC超级计算机提供了第一个分子动力学应用程序包,称为MDoC,该程序包是为QCD(量子色动力学)应用程序设计的。由于硬件架构(尤其是网络)的高度复杂性和独特性,其他分子模拟软件包将需要进行重大开发才能利用QCDOC功能。 MDoC是定制设计的,可以充分利用这种独特的架构优势。典型的生物分子系统具有数千甚至可能数百万个原子,其时标在微秒到毫秒之间。模拟它们需要使用超级计算机才能提供的大量计算能力。模拟生物分子最耗时的计算是静电相互作用的计算,通常通过使用Ewald算法来处理。我们还将介绍分子模拟所需的其他算法,特别是使用Melchionna修改后的Nose-Hoover方法的恒温和恒压算法。我们讨论了一种有效的QCDOC全聚集方法。由于QCDOC没有像BlueGene / L上的树形网络那样用于全局通信的特定于网络的算法,因此我们提出一种专门用于处理远程静电相互作用的算法。我们证明了QCDOC对于非QCD应用程序的可用性。由于快速的通信网络和PowerPC处理器的灵活性,当对具有约30,000个原子的蛋白质(包括溶剂)使用1024个处理器时,分子动力学仿真能够实现70%的并行效率。

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