中文摘要：

针对生物化学计算中采用量子化学理论计算蛋白质分子场所带来的巨大计算量的问题，搭建起一个GPU集群系统，用来加速计算基于量子化学的蛋白质分子场．该系统采用消息传递并行编程环境（MPI）连接集群各结点，以开放多线程OpenMP编程标准作为多核CPU编程环境，以CUDA语言作为GPU编程环境，提出并实现了集群系统结点中GPU和多核CPU协同计算的并行加速架构优化设计．在保持较高计算精度的前提下，结合MPI，OpenMP和CUDA混合编程模式，大大提高了系统的计算性能，并对不同体系和规模的蛋白质分子场模拟进行了计算分析．与相应的CPU集群、GPU单机和CPU单机计算方法对比，该GPU集群大幅度地提高了高分辨率复杂蛋白质分子场模拟的计算效率，比CPU集群的平均计算加速比提高了7．5倍．

英文摘要：

This paper reports a new GPU cluster system to meet the challenge of computing the protein molecular field by approaches of quantum chemistry. This system uses MPI （message passing interface） parallel programming environment to link all nodes of the cluster, adopts OpenMP （open multi-processing） as a multi-core CPU programming environment and CUDA as a GPU programming environment. We present an optimal parallel accelerating architecture to integrate multi-core CPU computation with GPU computation at each node of the GPU cluster system. By combining MPI, OpenMP and CUDA programming, it greatly speeds up the computation while maintaining high calculation precision. We tested the cluster system by calculating several protein molecular fields adhering to the theory of quantum chemistry. Compared with the previous approaches, such as CPU cluster, single GPU and single CPU, the proposed GPU cluster improves the calculation efficiency greatly. One example shows 7.5 times acceleration by employing the GPU cluster system in comparison with that by a corresponding CPU cluster. It is therefore capable of simulating the complex protein molecular field at high resolution.