GPU计算:突破制约计算化学发展的瓶颈  

GPU Computing: Overcoming the Bottleneck to Computational Chemistry Development

作　　者：刘松[1,2] 鲍建樟[3] 李长瑜 于建国[2] 

机构地区：[1]安徽师范大学化学与材料科学学院,安徽芜湖241000 [2]北京师范大学化学学院,北京100875 [3]北京师范大学系统科学学院,北京100875 [4]北京宏剑公司,北京100080

出　　处：《科研信息化技术与应用》2014年第2期73-81,共9页E-science Technology & Application

基　　金：国家自然科学基金(21073014)

摘　　要：在生物体系、材料科学和药物设计等领域,采用计算化学方法存在计算耗时长、计算精度难以提高、计算体系难以扩大等瓶颈问题。本文介绍了使用GPU加速的量子化学计算程序和分子动力学模拟程序的新进展,并着重对完全为GPU计算设计的量子化学计算程序TeraChem和分子动力学模拟程序ACEMD在算法、功能、性能和应用等方面做了详细的介绍,展望了GPU技术在计算化学领域中新的机遇和挑战。As well known, the applications of computational chemistry in biology, materials science, and drug design, have the problems of time-consuming, lower accuracy and bad scalability of the calculation system, and the bottleneck is hard to break through. In this paper, we reviewed the progress of GPU algorithms and methods utilized in computational chemistry. Then we introduced the GPU quantum chemical software TeraChem and MD package ACEMD in detail, which can speed up dozens of times comparing the CPU packages. In the end, an overlook of the opportunities and challenges for GPU technology in computational chemistry was presented.

关 键 词：GPU 量子化学 分子动力学 TeraChem ACEMD 

分 类 号：G642.0[文化科学—高等教育学]