Linear least squares solutions by Householder transformations with column pivoting on a parallel machine

Author

Ma, Sangback

Author_Institution

Dept. of Comput. Sci., Hanyang Univ., Ansan, South Korea

fYear

1997

fDate

28 Apr-2 May 1997

Firstpage

134

Lastpage

136

Abstract

Using Householder transformations to solve linear least squares problems is very efficient. We have found a computational form especially suitable for BBN Butterfly, a parallel computer. We modified the algorithm for parallel processing using various constructs from the BBN Mach 1000 system. We were able to obtain a speed-up of 3 with 4 processors, and 3.83 with 6 processors for M=1000, and N=200, which means Ep=75% and 64% CPU utilization. We conclude that our algorithm realized a reasonably good speedup with a moderate number of processors

Keywords

least squares approximations; mathematics computing; matrix algebra; parallel algorithms; parallel machines; shared memory systems; software performance evaluation; vectors; BBN Butterfly; BBN Mach 1000 system; CPU utilization; Householder transformations; column pivoting; linear least squares problems; matrices; parallel algorithm; parallel computer; parallel machine; parallel processing; shared memory system; speed-up; vectors; Artificial intelligence; Computer science; Concurrent computing; Grain size; Least squares methods; Parallel machines; Parallel processing; Symmetric matrices;

fLanguage

English

Publisher

ieee

Conference_Titel

High Performance Computing on the Information Superhighway, 1997. HPC Asia '97

Conference_Location

Seoul

Print_ISBN

0-8186-7901-8

Type

conf

DOI

10.1109/HPC.1997.592136

Filename

592136