基于GPU的生物大分子计算平台的构建与优化

Construction and Optimization of GPU-Based Computing Platform for Biological Macromolecules

通过生物大分子计算平台对分子动力学的模拟运算,计算值可以描述分子的运动轨迹,从而揭示原子间的内在关系,但这一过程较为复杂且昂贵。作者研究了较低投入下大分子计算平台的构建与优化,在并行GPU条件和CUDA体系下使用AMBER软件包完成分子动力学的模拟运算。通过对两个大分子体系的运算,实验结果表明计算平台达到了60倍的计算加速,实现了较高的加速性能,完成了高性价比的高性能计算平台的建设。

Biological macromolecules by molecular dynamics computing platform for analog operation,calculated value can describe the trajectory of molecules,which reveals the intrinsic relationship between atoms,but the process is more complicated and expensive. In this paper,the construction and optimization to achieve lower investment macromolecules computing platforms, using a software package AMBER molecular dynamics simulation runs in parallel GPU and CUDA system conditions. Through the operation of two large molecules, experimental results show that the computing platform reached 60 times the computing speed up to achieve a high acceleration performance, completed the construction of cost-effective, high-performance computing platform.