Self-adapting Linear Algebra algorithms and software

James Demmel; Jack Dongarra; Victor Eijkhout; Erika Fuentes; Antoine Petitet; Richard Vuduc; R. Clint Whaley; Katherine Yelick

doi:10.1109/JPROC.2004.840848

Self-adapting Linear Algebra algorithms and software

James Demmel, Jack Dongarra, Victor Eijkhout, Erika Fuentes, Antoine Petitet, Richard Vuduc, R. Clint Whaley, Katherine Yelick

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

Abstract

One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept. © 2005 IEEE.

Original language	English
Pages (from-to)	293-311
Number of pages	18
Journal	Institute of Electrical and Electronics Engineers. Proceedings
Volume	93
Issue number	2
DOIs	https://doi.org/10.1109/JPROC.2004.840848
Publication status	Published - Feb 2005

Keywords

Adaptive methods
Basic Linear Algebra Subprograms (BLAS)
Dense kernels
Iterative methods
Linear systems
Matrix-matrix product
Matrix-vector product
Performance optimization
Preconditioners
Sparse kernels

Access to Document

10.1109/JPROC.2004.840848

Cite this

@article{91adf6844dc8464ea2d3e07eabc86a04,

title = "Self-adapting Linear Algebra algorithms and software",

abstract = "One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept. {\textcopyright} 2005 IEEE.",

keywords = "Adaptive methods, Basic Linear Algebra Subprograms (BLAS), Dense kernels, Iterative methods, Linear systems, Matrix-matrix product, Matrix-vector product, Performance optimization, Preconditioners, Sparse kernels",

author = "James Demmel and Jack Dongarra and Victor Eijkhout and Erika Fuentes and Antoine Petitet and Richard Vuduc and Whaley, {R. Clint} and Katherine Yelick",

year = "2005",

month = feb,

doi = "10.1109/JPROC.2004.840848",

language = "English",

volume = "93",

pages = "293--311",

journal = "Institute of Electrical and Electronics Engineers. Proceedings ",

issn = "0018-9219",

publisher = "IEEE",

number = "2",

}

TY - JOUR

T1 - Self-adapting Linear Algebra algorithms and software

AU - Demmel, James

AU - Dongarra, Jack

AU - Eijkhout, Victor

AU - Fuentes, Erika

AU - Petitet, Antoine

AU - Vuduc, Richard

AU - Whaley, R. Clint

AU - Yelick, Katherine

PY - 2005/2

Y1 - 2005/2

N2 - One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept. © 2005 IEEE.

AB - One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept. © 2005 IEEE.

KW - Adaptive methods

KW - Basic Linear Algebra Subprograms (BLAS)

KW - Dense kernels

KW - Iterative methods

KW - Linear systems

KW - Matrix-matrix product

KW - Matrix-vector product

KW - Performance optimization

KW - Preconditioners

KW - Sparse kernels

U2 - 10.1109/JPROC.2004.840848

DO - 10.1109/JPROC.2004.840848

M3 - Article

SN - 0018-9219

VL - 93

SP - 293

EP - 311

JO - Institute of Electrical and Electronics Engineers. Proceedings

JF - Institute of Electrical and Electronics Engineers. Proceedings

IS - 2

ER -

Self-adapting Linear Algebra algorithms and software

Abstract

Keywords

Access to Document

Fingerprint

Cite this