Investigation on the use of GPGPU for fast sparse matrix factorization

Pao Yue-kong Library Electronic Theses Database

Investigation on the use of GPGPU for fast sparse matrix factorization

 

Author: Tian, Ye
Title: Investigation on the use of GPGPU for fast sparse matrix factorization
Degree: M.Sc.
Year: 2010
Subject: Graphics processing units -- Programming.
Parallel programming (Computer science)
Computer algorithms.
Hong Kong Polytechnic University -- Dissertations
Department: Dept. of Electrical Engineering
Pages: v, 70 leaves : ill. (some col.) ; 30 cm.
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b2463053
URI: http://theses.lib.polyu.edu.hk/handle/200/6456
Abstract: In power system analysis, solution for network equations is frequently encountered. As the system size becomes larger, time consumed in network solution becomes a dominant factor in the overall time cost. One distinct and most important feature of the network admittance matrix is that it is highly sparse. Over the years, different methods have been developed to exploit the sparsity of matrix. LU factorization is a fundamental process in the solution for the network linear equations. One technique to improve the LU factorization time is parallel computation, with which data processing can be divided into different tasks and implemented simultaneously. However, up to now, efficiency of parallel computation algorithm implemented on multi-processor systems is adversely affected by the data communication latency between processors. In this paper, parallel computing power of the contemporary Graphic Processing Units (GPU) is exploited and sparse technique for LU factorization implemented on GPU is designed. Parallel LU factorization algorithm of partitioned matrix is proposed. To better cater for dense data formats suitability of GPU, Cholesky factorization with a supernodal approach is implemented where supernodes are computed on GPU. All algorithms are implemented on the Computer Unified Device Architecture (CUDA) of the NVIDIA GPU. Performance and factors affecting performance are studied and analyzed. Revisions are made to some existing algorithms and new ideas are proposed for future study.

Files in this item

Files Size Format
b24630536.pdf 2.741Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information