Development of high performance computation system with great precision for molecular dynamics simulations

分子动力学模拟为研究分子系统的结构和动力学提供了有价值的手段。由于它需要大量的计算资源，特别是当它被应用到大型生物系统，它提出了具有挑战性的问题，计算技术。虽然有一些快速算法被开发出来以减少计算时间，但这些算法都遇到了一定的困难。在这项工作中，我首先重新制定了快速多极子方法（FMM），以进一步减少计算时间和内存需求。其次，我开发了两个可靠的算法来实现FMM框架中的周期性。一种方法采用Ewald样技术，另一种方法使用细胞的宏观层次结构。由于包括FMM在内的每一种快速算法都采用了某种近似，因此这些算法在某些情况下会受到不准确性的影响。在这项工作中，我发现，使用一个特殊用途的机器的任务涉及FMM提高精度和计算速度。这种具有新颖算法的计算系统实际上是在这项工作中开发的。该系统的性能比普通工作站高几百倍。

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. In this work, I first reformulated the fast multipole method (FMM) for further decreasing the computation time and memory requirements. Secondly, I developed two reliable algorithms for accomplishing periodicity in the FMM framework. One method employs the Ewald-like technique, and the other uses a macroscopic hierarchical structure of cells. These methods developed in this work enable reliable simulations of biological macromolecules.Since every fast algorithm including FMM employs a certain kind of approximation, the algorithms suffer from inaccuracy in some situation. In this work, I found that use of a special-purpose machine for a task involved in FMM improve both accuracy and computation speed. Such a computation system with novel algorithms was actually developed in this work. The performance of the system was confirmed to be several hundred times higher than those of ordinary workstations.