Fast algorithm/ hardware joint acceleration for molecular dynamics simulations and its efficacy confirmation

分子动力学模拟快速算法/硬件联合加速及其有效性验证

基本信息

批准号：
13680743
负责人：
AMISAKI Takashi
金额：
$ 2.24万
依托单位：
Tottori University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2001
资助国家：
日本
起止时间：
2001 至 2002
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-13680743/
关键词：
molecular dynamics proteins fast multipole methods special-purpose machine parallel computing cluster computing computation accuracy MD-Engine SHAKE 専用ハードウェア

项目摘要

Molecular dynamics simulation provides valuable means to study the structure and dynamics of molecular systems. Since it demands vast computational resources, particularly when it is applied to large biological systems, it presents challenging problems to computational technology. Although several fast algorithms were developed for reducing the computation time, these algorithms encounter certain difficulties.To overcome this problem, in this work, I developed algorithms that realized efficient cooperation of dedicated computational board (MD-Engine II) with fast algorithms, such as fast multipole method (FMM). The algorithms were implemented on a PC, into which the computation boards were plugged, to acutally build up a specialized computation unit for MD simulations. I then developed a cluster of the four computation units to realize a high performace system for accurate MD calculation at low cost. Using the computation system, 1MD step of a protein-water system (fifty thousand atoms) can be completed in 1.6s. In addition, the results of a series of 50 ps MD simulations of a protein-water system (fifty thousand atoms) revealed that a more stringent setting of accuracy in FMM computation, as compared with those previously reported, was required for accurate simulations over long time periods.

分子动力学模拟提供了研究分子系统的结构和动力学的宝贵手段。由于它需要大量的计算资源，尤其是将其应用于大型生物系统时，它给计算技术带来了挑战性问题。尽管开发了几种快速算法来减少计算时间，但这些算法遇到了某些困难。要克服这一问题，在这项工作中，我开发了算法，这些算法实现了专用计算板（MD-Engine II）与快速算法的有效合作，例如快速多层方法（FMM）。这些算法是在PC上实现的，该算法插入了计算板上，以实现用于MD模拟的专门计算单元。然后，我开发了一个四个计算单元的群集，以实现高性能系统，以低成本以准确的MD计算。使用计算系统，可以在1.6s中完成蛋白质水系统（五万个原子）的1MD步骤。此外，与先前报道的那些相比，在长时间内，需要精确模拟需要一系列蛋白质水系统（五万个原子）的50 PS MD模拟（五万个原子）的结果。