COMPUTATIONAL FACILITY

MDScope建模环境NAMD的仿真组件 [19]继续在功能和速度上进行改进*。 最新版本NAMD2(完全重写了NAMD的大部分内容 代码)，即将公开发布。NAMD2比任何 上一版本在实现NAMD的可修改性设计目标方面， 速度和并行性能。NAMD2比 原始版本，如上一个中添加的功能所示 年。该程序现在支持周期性边界条件，使用 要么切断静电，要么完全Ewald静电。它允许 使用三重时间步长方案的更大模拟步长，这可以 允许某些模拟以原始速度的两倍运行 方法。NAMD2有一个基于TCL的原型脚本接口， 允许用户更改模拟的某些部分，而不必 请看一下程序代码。它现在还雇佣了DPME 完全静电学的算法，由我们的合作者在 杜克大学。NAMD2支持大量并行机， 包括工作站集群、SGI Origin 200和Origin 2000、IBM SP3、Convex Example和Cray T3E。新的 程序的结构表现出改进的顺序速度，并且 在几十到数百台计算机上实现更好的并行性能[22] 处理器。改进的算法和并行效率，以及 大量处理器的供应导致了一笔订单 提高了模拟速度的数量级。

The simulation component of the MDScope modeling environment, NAMD [19], continues to undergo improvement in both features and speed*. The latest version, NAMD2 (a complete rewrite of much of the NAMD code), is nearing a public release. NAMD2 goes further than any previous release in achieving the NAMD design goals of modifiability, speed, and parallel performance. NAMD2 is more modifiable than the original version, as demonstrated by the features added in the last year. The program now supports periodic boundary conditions, using either cutoff electrostatics or full Ewald electrostatics. It allows larger simulation steps using a triple-time-stepping scheme, which can allow some simulations to run at double the speed of the original method. NAMD2 has a prototype TCL-based scripting interface that permits users to alter some parts of the simulation without having to look at the program code at all. It now also employs the DPME algorithm for full electrostatics, developed by our collaborators at Duke University. NAMD2 supports a large number of parallel machines, including clusters of workstations, the SGI Origin 200 and Origin 2000, the IBM SP3, the Convex Exemplar, and the Cray T3E. The new structure of the program exhibits improved sequential speed, and better parallel performance [22] on machines with dozens to hundreds of processors. Improved algorithms and parallel efficiency, and the availability of large numbers of processor have resulted in an order of magnitude increase in the speed of simulations.