高性能应用软件并行优化策略的研究 |
| |
作者姓名: | 贾伟乐 史小冬 吕海峰 |
| |
作者单位: | 1. 北京北龙超级云计算有限责任公司, 北京 100190;
2. 中国科学院计算机网络信息中心, 北京 100190;
3. 中国科学院大学, 北京 100190 |
| |
摘 要: | 多核时代的来临对现有的应用软件提出了严重挑战,串行代码难以充分发挥硬件资源的性能;软件的并行优化成为亟待解决的重要问题。本文综合了 MPI,OpenMP,众核编程模型 CUDA 三个编程模型进行研究,讨论了适用于不同软件并行优化的方法,提出了适用于企业级应用的软件并行优化策略,最后总结和展望了软件并行优化的挑战和前景。
|
关 键 词: | 并行优化 MPI OpenMP CUDA GPU |
收稿时间: | 2013-05-24 |
On the Parallelization Optimization Strategy for High Performance Computing Software |
| |
Authors: | Jia Weile Shi Xiaodong L Haifeng |
| |
Affiliation: | 1. Beijing Beilong Super Cloud Computing Co. Ltd, Beijing 100190, China;
2. Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China;
3. University of Chinese Academy of Sciences, Beijing 100190, China |
| |
Abstract: | With the arrival of the multi-core era, traditional software can barely utilize the peak performance of the hardware. Parallelization and optimization of the industrial code is a tough problem the HPC community is facing. In this paper, we present our parallel im-plementation using MPI, OpenMP and CUDA programming models. Different methods, implementation skills and optimization strategies are introduced. At the end, the challenge and the vision of the future work of our optimization strategy are discussed |
| |
Keywords: | parallel optimization MPI OpenMP CUDA GPU |
|
| 点击此处可从《》浏览原始摘要信息 |
|
点击此处可从《》下载全文 |
|