期刊文献+

基于CUDA的医学影像数据处理工作站的配置方法 预览

The Configuration of Medical Imaging Data Analysis Workstation Based on CUDA
在线阅读 免费下载
收藏 分享 导出
摘要 目的:以个人电脑为平台,结合CUDA(Compute Unified Device Architecture,统一计算设备架构)以显著提高个人电脑的计算能力,使其能够承担大运算量医学影像数据处理任务。方法:CUDA是NVIDIA公司推出的一款运算模型,把GPU多线程并行性能引入除图像显示之外的领域,可以大幅提升个人电脑的运算能力。在装有Windows XP的电脑上安装支持CUDA的NVIDIA显卡,然后依次下载安装CUDA Driver,toolkit,SDK,Visual Studio及CUDA VS wizardx32软件。让CPU和GPU协同工作,把程序中可以并行处理的部分并行化,使原来在CPU上只能顺序处理的大量循环计算,可以在GPU上以大量线程并行处理的方式一次计算完成。结果:在配备支持CUDA的NVIDIA显卡的个人电脑上可以安装CUDA,但配置比较复杂,安装完成通过测试后可以利用显卡中大量的流处理器进行并行计算。结论:个人电脑安装CUDA后可以利用GPU中大量的流处理器进行并行计算,将个人电脑转变为一台高性价比的医学影像数据并行处理工作站。 Objective: To implement the CUDA(Compute Unified Device Architecture) in conventional personal computers,in order to improve the computing power of personal computer and enable it taking on the task of disposing a mass of medical image data.Methods: CUDA is a calculation architecture developed by NVIDIA Corporation,which introduces the multithreading calculation to many new fields besides imaging display and upgrades the calculation power of personal computer.In the beginning,we install a NVIDID display card supporting CUDA in the computer with Windows XP,then,download and install CUDA Driver,toolkit,SDK,Visual Studio and CUDA VS wizard in order,let CPU work together with GPU,and parallelize the part of program which can be managed paralleling.These steps will dispose operations once with lots of threads on GPU,instead of repeating them again and again on CPU.Results: CUDA can be installed on the computer with NVIDIA display card supporting CUDA,while the configuration is complex.After finishing installation and passing the test,a large number of stream processors in GPU may be used in the process of medical imaging data analysis.Conclusions: A conventional personal computer may be used as a cost-effective parallel medical imaging data analysis workstation after implement CUDA in it.
作者 王飞 高嵩 WANG Fei,GAO Song(Medical Imaging Physics and Engineering Laboratory,Peking University Health Science Center,Beijing 100083,China)
出处 《中国医学物理学杂志》 CSCD 2011年第6期 3004-3006,3011,共4页 Chinese Journal of Medical Physics
基金 国家自然科学基金资助项目(No.81171330) 北京市自然科学基金资助项目(No.7102102) 东莞市高等院校科研机构科技发展项目(No.200910815252)
关键词 统一计算设备架构 CUDA 并行计算 compute unified device architecture CUDA multithread calculation
  • 相关文献

参考文献4

  • 1Y.Okitsu,F.Ino,K.Hagihara.High-performance cone beam recon-struction using CUDA compatible GPUs[].Parallel Computing.2010 被引量:1
  • 2.NVIDIA Fermi Compute Architecture Whitepaper[]..2009 被引量:1
  • 3.NVIDIA CUDA C Programming Guide,Version 3.2[]..2010 被引量:1
  • 4NO L P B,,WALCZAK A M,XU J H,et al.GPU-basedcone beamcomputed tomography[].Computer Methods andPrograms in Biomedicine.2010 被引量:1
投稿分析

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 意见反馈