Mixed precision iterative refinement techniques for the solution of dense linear systems

Alfredo Buttari; Jack Dongarra; Julie Langou; Julien Langou; Piotr Luszczek; Jakub Kurzak

doi:10.1177/1094342007084026

Mixed precision iterative refinement techniques for the solution of dense linear systems

Alfredo Buttari, Jack Dongarra, Julie Langou, Julien Langou, Piotr Luszczek, Jakub Kurzak

Computer Science and Mathematics Div

Research output: Contribution to journal › Article › peer-review

100 Scopus citations

Abstract

By using a combination of 32-bit and 64-bit floating point arithmetic, the performance of many dense and sparse linear algebra algorithms can be significantly enhanced while maintaining the 64-bit accuracy of the resulting solution. The approach presented here can apply not only to conventional processors but also to exotic technologies such as Field Programmable Gate Arrays (FPGA), Graphical Processing Units (GPU), and the Cell BE processor. Results on modern processor architectures and the Cell BE are presented.

Original language	English
Pages (from-to)	457-466
Number of pages	10
Journal	International Journal of High Performance Computing Applications
Volume	21
Issue number	4
DOIs	https://doi.org/10.1177/1094342007084026
State	Published - Dec 2007

Keywords

Cholesky factorization
Iterative
LU
Mixed-precision
Refinement

Access to Document

10.1177/1094342007084026

Cite this

@article{98b13f0ec86242369464e1e6c5f51e14,

title = "Mixed precision iterative refinement techniques for the solution of dense linear systems",

abstract = "By using a combination of 32-bit and 64-bit floating point arithmetic, the performance of many dense and sparse linear algebra algorithms can be significantly enhanced while maintaining the 64-bit accuracy of the resulting solution. The approach presented here can apply not only to conventional processors but also to exotic technologies such as Field Programmable Gate Arrays (FPGA), Graphical Processing Units (GPU), and the Cell BE processor. Results on modern processor architectures and the Cell BE are presented.",

keywords = "Cholesky factorization, Iterative, LU, Mixed-precision, Refinement",

author = "Alfredo Buttari and Jack Dongarra and Julie Langou and Julien Langou and Piotr Luszczek and Jakub Kurzak",

year = "2007",

month = dec,

doi = "10.1177/1094342007084026",

language = "English",

volume = "21",

pages = "457--466",

journal = "International Journal of High Performance Computing Applications",

issn = "1094-3420",

publisher = "SAGE Publications",

number = "4",

}

TY - JOUR

T1 - Mixed precision iterative refinement techniques for the solution of dense linear systems

AU - Buttari, Alfredo

AU - Dongarra, Jack

AU - Langou, Julie

AU - Langou, Julien

AU - Luszczek, Piotr

AU - Kurzak, Jakub

PY - 2007/12

Y1 - 2007/12

N2 - By using a combination of 32-bit and 64-bit floating point arithmetic, the performance of many dense and sparse linear algebra algorithms can be significantly enhanced while maintaining the 64-bit accuracy of the resulting solution. The approach presented here can apply not only to conventional processors but also to exotic technologies such as Field Programmable Gate Arrays (FPGA), Graphical Processing Units (GPU), and the Cell BE processor. Results on modern processor architectures and the Cell BE are presented.

AB - By using a combination of 32-bit and 64-bit floating point arithmetic, the performance of many dense and sparse linear algebra algorithms can be significantly enhanced while maintaining the 64-bit accuracy of the resulting solution. The approach presented here can apply not only to conventional processors but also to exotic technologies such as Field Programmable Gate Arrays (FPGA), Graphical Processing Units (GPU), and the Cell BE processor. Results on modern processor architectures and the Cell BE are presented.

KW - Cholesky factorization

KW - Iterative

KW - LU

KW - Mixed-precision

KW - Refinement

UR - http://www.scopus.com/inward/record.url?scp=35548933706&partnerID=8YFLogxK

U2 - 10.1177/1094342007084026

DO - 10.1177/1094342007084026

M3 - Article

AN - SCOPUS:35548933706

SN - 1094-3420

VL - 21

SP - 457

EP - 466

JO - International Journal of High Performance Computing Applications

JF - International Journal of High Performance Computing Applications

IS - 4

ER -

Mixed precision iterative refinement techniques for the solution of dense linear systems

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this