- 1、本文档共42页,可阅读全部内容。
- 2、有哪些信誉好的足球投注网站(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
- 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
PAGE
摘 要
高斯消去法是求解线性方程组的最基本的方法之一。为了充分利用GPU (Graphics Processing Unit,图形处理器)的并行处理能力,本文改进了高斯列主元消去法的实现过程,从而提高了求解线性方程组的速度。并研究了在不同方程组阶数下,GPU对这此算法的加速效果。NVIDIA新近推出的GPU计算平台采用矩阵型的计算架构,对处理大型矩阵具有极大的优势,且相对CPU有着更高的算法可并行性和计算效率。本文力图基于GPU的CUDA开发环境,利用GP-GPU的计算特性实现求解线性方程组,以提高算法的运行效率。最后,本文用C语言实现了高斯列主元消去算法求解线性方程组的基本过程,并分别在NVIDIA GPU并行计算平台和Intel CPU计算平台上加以运行,同时进行了两种计算平台上算法实现的性能比较。
关键词:求解线性方程组;高斯消去法;GPU;CUDA;并行计算
Abstract
Gaussian elimination method is one of the most basic methods for solving linear equations. In order to take full advantage of GPU (Graphics Processing Unit, GPU) of the parallel processing capability, this paper improved Gaussian PCA out of the realization of the process of elimination, resulting in improved system of linear equations to solve the speed. And studied the different equations in a few bands, GPU accelerated the effect of this algorithm. NVIDIA recently introduced the use of GPU computing platform for the calculation of matrix-type structure, to deal with large-scale matrix has great advantages, and the relative CPU algorithm has higher computational efficiency and parallelism. This article seeks to CUDA for GPU-based development environment, the use of GP-GPU computing features to achieve the solution of linear equations in order to improve the efficiency of algorithm. Finally, using C language realization of the Gaussian elimination algorithm PCA out of linear equations to solve the basic process and NVIDIA GPU in parallel computing platform and Intel CPU computing platform to be run, at the same time on two types of computing platforms algorithm Performance Comparison of the achievement.
Key?words: Solving?linear?equations;?Gaussian?elimination?method; GPU; CUDA; Parallel?Computing
目 录
TOC \o 1-3 \h \z \u HYPERLINK \l _Toc231887905 第一章 绪论 PAGEREF _Toc231887905 \h 1
HYPERLINK \l _Toc231887906 1.1 引言 PAGEREF _Toc231887906 \h 1
HYPERLINK \l _Toc231887907 1.2 论文研究背景 PAGEREF _Toc231887907 \h 1
HYPERLINK \l _Toc231887908 1
文档评论(0)