Automatic optimisation of parallel linear algebra routines in systems with variable load

J. Cuenca; D. Giménez; J. González; J. Dongarra; K. Roche

doi:10.1109/EMPDP.2003.1183618

Automatic optimisation of parallel linear algebra routines in systems with variable load

J. Cuenca, D. Giménez, J. González, J. Dongarra, K. Roche

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

12 Scopus citations

Abstract

The architecture of an automatically tuned linear algebra library proposed in previous works is extended in order to adapt itself to platforms where both the CPU load and the network traffic vary. During the installation process in a system, the linear algebra routines will be tuned automatically to the system conditions: hardware characteristics and basic libraries used in the linear algebra routines. At run-time the parameters that define the system characteristics are adjusted to the actual load of the platform. The design methodology is analysed with a block LU factorisation. Variants for sequential and parallel versions of this routine on a logical rectangular mesh of processors are considered. The behavior of the algorithm is studied with message-passing, using MPI on a cluster of PCs. The experiments show that the configurable parameters of the linear algebra routines can be adjusted during the run-time process despite the variability of the environment.

Original language	English
Title of host publication	Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003
Editors	Andrea Clematis
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	409-416
Number of pages	8
ISBN (Electronic)	0769518753, 9780769518756
DOIs	https://doi.org/10.1109/EMPDP.2003.1183618
State	Published - 2003
Externally published	Yes
Event	11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003 - Genova, Italy Duration: Feb 5 2003 → Feb 7 2003

Publication series

Name	Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003

Conference

Conference	11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003
Country/Territory	Italy
City	Genova
Period	02/5/03 → 02/7/03

Keywords

Clustering algorithms
Computer architecture
Computer science
Hardware
Libraries
Lifting equipment
Linear algebra
Runtime
Supercomputers
Telecommunication traffic

Access to Document

10.1109/EMPDP.2003.1183618

Cite this

Cuenca, J., Giménez, D., González, J., Dongarra, J., & Roche, K. (2003). Automatic optimisation of parallel linear algebra routines in systems with variable load. In A. Clematis (Ed.), Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003 (pp. 409-416). Article 1183618 (Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/EMPDP.2003.1183618

Cuenca, J. ; Giménez, D. ; González, J. et al. / Automatic optimisation of parallel linear algebra routines in systems with variable load. Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003. editor / Andrea Clematis. Institute of Electrical and Electronics Engineers Inc., 2003. pp. 409-416 (Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003).

@inproceedings{dcd1c13e089346868121703b8e89212d,

title = "Automatic optimisation of parallel linear algebra routines in systems with variable load",

abstract = "The architecture of an automatically tuned linear algebra library proposed in previous works is extended in order to adapt itself to platforms where both the CPU load and the network traffic vary. During the installation process in a system, the linear algebra routines will be tuned automatically to the system conditions: hardware characteristics and basic libraries used in the linear algebra routines. At run-time the parameters that define the system characteristics are adjusted to the actual load of the platform. The design methodology is analysed with a block LU factorisation. Variants for sequential and parallel versions of this routine on a logical rectangular mesh of processors are considered. The behavior of the algorithm is studied with message-passing, using MPI on a cluster of PCs. The experiments show that the configurable parameters of the linear algebra routines can be adjusted during the run-time process despite the variability of the environment.",

keywords = "Clustering algorithms, Computer architecture, Computer science, Hardware, Libraries, Lifting equipment, Linear algebra, Runtime, Supercomputers, Telecommunication traffic",

author = "J. Cuenca and D. Gim{\'e}nez and J. Gonz{\'a}lez and J. Dongarra and K. Roche",

note = "Publisher Copyright: {\textcopyright} 2003 IEEE.; 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003 ; Conference date: 05-02-2003 Through 07-02-2003",

year = "2003",

doi = "10.1109/EMPDP.2003.1183618",

language = "English",

series = "Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "409--416",

editor = "Andrea Clematis",

booktitle = "Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003",

}

Cuenca, J, Giménez, D, González, J, Dongarra, J & Roche, K 2003, Automatic optimisation of parallel linear algebra routines in systems with variable load. in A Clematis (ed.), Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003., 1183618, Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003, Institute of Electrical and Electronics Engineers Inc., pp. 409-416, 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003, Genova, Italy, 02/5/03. https://doi.org/10.1109/EMPDP.2003.1183618

Automatic optimisation of parallel linear algebra routines in systems with variable load. / Cuenca, J.; Giménez, D.; González, J. et al.
Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003. ed. / Andrea Clematis. Institute of Electrical and Electronics Engineers Inc., 2003. p. 409-416 1183618 (Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Automatic optimisation of parallel linear algebra routines in systems with variable load

AU - Cuenca, J.

AU - Giménez, D.

AU - González, J.

AU - Dongarra, J.

AU - Roche, K.

PY - 2003

Y1 - 2003

N2 - The architecture of an automatically tuned linear algebra library proposed in previous works is extended in order to adapt itself to platforms where both the CPU load and the network traffic vary. During the installation process in a system, the linear algebra routines will be tuned automatically to the system conditions: hardware characteristics and basic libraries used in the linear algebra routines. At run-time the parameters that define the system characteristics are adjusted to the actual load of the platform. The design methodology is analysed with a block LU factorisation. Variants for sequential and parallel versions of this routine on a logical rectangular mesh of processors are considered. The behavior of the algorithm is studied with message-passing, using MPI on a cluster of PCs. The experiments show that the configurable parameters of the linear algebra routines can be adjusted during the run-time process despite the variability of the environment.

AB - The architecture of an automatically tuned linear algebra library proposed in previous works is extended in order to adapt itself to platforms where both the CPU load and the network traffic vary. During the installation process in a system, the linear algebra routines will be tuned automatically to the system conditions: hardware characteristics and basic libraries used in the linear algebra routines. At run-time the parameters that define the system characteristics are adjusted to the actual load of the platform. The design methodology is analysed with a block LU factorisation. Variants for sequential and parallel versions of this routine on a logical rectangular mesh of processors are considered. The behavior of the algorithm is studied with message-passing, using MPI on a cluster of PCs. The experiments show that the configurable parameters of the linear algebra routines can be adjusted during the run-time process despite the variability of the environment.

KW - Clustering algorithms

KW - Computer architecture

KW - Computer science

KW - Hardware

KW - Libraries

KW - Lifting equipment

KW - Linear algebra

KW - Runtime

KW - Supercomputers

KW - Telecommunication traffic

UR - http://www.scopus.com/inward/record.url?scp=84902775394&partnerID=8YFLogxK

U2 - 10.1109/EMPDP.2003.1183618

DO - 10.1109/EMPDP.2003.1183618

M3 - Conference contribution

AN - SCOPUS:84902775394

T3 - Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003

SP - 409

EP - 416

BT - Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003

A2 - Clematis, Andrea

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003

Y2 - 5 February 2003 through 7 February 2003

ER -

Cuenca J, Giménez D, González J, Dongarra J, Roche K. Automatic optimisation of parallel linear algebra routines in systems with variable load. In Clematis A, editor, Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003. Institute of Electrical and Electronics Engineers Inc. 2003. p. 409-416. 1183618. (Proceedings - 11th Euromicro Conference on Parallel, Distributed and Network-Based Processing, Euro-PDP 2003). doi: 10.1109/EMPDP.2003.1183618

Automatic optimisation of parallel linear algebra routines in systems with variable load

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this