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体系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，特别是当它应用于大型生物系统时，它对计算技术提出了挑战性的问题。为了减少计算时间，人们开发了几种快速算法，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方法（FMM），以进一步减少计算时间和内存需求。其次，开发了两种可靠的算法来实现FMM框架下的周期性。一种方法采用了类似伊瓦尔德的技术，而另一种方法则使用了细胞的宏观层次结构。在这项工作中开发的这些方法能够可靠地模拟生物大分子。由于包括FMM在内的每一种快速算法都采用了某种近似，因此算法在某些情况下存在不准确性。在这项工作中，我发现使用专用机器来完成FMM中涉及的任务可以提高精度和计算速度。本研究实际开发了一个具有新颖算法的计算系统。系统的性能比普通工作站提高了几百倍。

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.