期刊文献+

CPU+GPU异构模式下并行计算效率研究 被引量:5

Research on Efficiency of Parallel Computing in CPU+GPU Hybrid Mode
在线阅读 下载PDF
导出
摘要 CPU+GPU的异构模式由于比传统的超算架构更加便宜和更加环保、低碳,所以得到了越来越多的关注,在HPC的Top500中也渐渐出现了异构模式的身影。然而异构模式下的并行效率过低也是个既定的事实。本文从异构模式及GPU之间并行调度的原理出发,以Linpack测试效率为例来展开异构模式下的并行计算效率研究,并给出相应结论。 CPU+GPU hybrid mode,comparing to traditional supercomputing architectures,is more cheaper and more environmentally friendly,low-carbon.Thus the people pay more attention on it.There are more and more systems which rank on World HPC Top500 using CPU+GPU hybrid mode.However,the efficiency of heterogeneous parallel mode is too low.This article based on the principle of parallel schedule of hybrid mode,takes Linpack test as an example to explain the method of getting a higher efficiency of parallel computing.
出处 《计算机与现代化》 2012年第5期185-188,共4页 Computer and Modernization
关键词 LINPACK 异构模式 GPGPU 超级计算 HPC 并行计算 Linpack hybrid mode GPGPU supercomputing HPC parallel computing
  • 相关文献

参考文献14

  • 1太平洋电脑网.IBM再次摘得世界最快的超级计算机桂冠[EB/OL].http://news.peonline.corn.c./giyj/0806/1329027.html.2008-06-19.
  • 2Wikipedia. IBM Roadrunner[ EB/OL]. http ://en. wikipe- dia. org/wiki/IBM_Roadrunner, 2011-09-16.
  • 3商用及工作站.NVIDIA:Tesla+CUDA是最好的HPC工具[EB/OL].http://epc.z01.tom.cn/169/1697028.ht-ml,2010-03-15.
  • 4Randy Femando. GPGPU: General-Purpose Computation on GPUs [ EB/OL]. http://http, download, nvidia, corn/ developer/presentations/20OS/I3D/I3D 05 GPGPU. pdf, 2012-01-09.
  • 5Wikipedia. Linpack [ EB/OL]. http://en, wikipedia, org/ wiki/LINPACK, 2011-11-02.
  • 6SC. The SC Conference Series[ EB/OL]. http ://www. su- percomp, org/about, php, 200t-12-28.
  • 7国家超级计算天津中心.“天河一号”(TH.1A)超级计算机[EB/OL].http://www.nscc-Ij.gov.cn/resources/re-sources_2.asp,2011-11-12.
  • 8博客园.CUDA并行计算框架(一)概念相关、内容比较抽象[EB/OL].http://www.cnblogs.com/dubing/archive/2011/10/10/2085721.html,2011-10-10.
  • 9Nvidia. NvidiaTesla[EB/OL]. http://www, nvidia.cn/ob- ject/tesla_computing_solutions_cn, html, 2012-01-09.
  • 10Nvidia. Whitepaper-NVIDIA' s Next Generation CUDATM Compute Architecture: FermiTM [ EB/OL]. http://www. nvidia, com/content/PDF/fermi _ white _ papers/NVIDIA _ Fermi _ Compute _ Architecture _ Whitepaper. pdf, 2012- 01-09.

共引文献2

同被引文献35

引证文献5

二级引证文献2

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部