Performance model for parallel mathematical libraries based on historical knowledgebase

I. Salawdeh; E. César; A. Morajko; T. Margalef; E. Luque

doi:10.1007/978-3-540-85451-7_13

Performance model for parallel mathematical libraries based on historical knowledgebase

I. Salawdeh^*, E. César, A. Morajko, T. Margalef, E. Luque

^*Autor corresponent d’aquest treball

Departament d'Arquitectura de Computadors i Sistemes Operatius

Producció científica: Contribució a revista › Article › Recerca › Avaluat per experts

1 Citació (Scopus)

Resum

Scientific and mathematical parallel libraries offer a high level of abstraction to programmers. However, it is still difficult to select the proper parameters and algorithms to maximize the application performance. This work proposes a performance model for dynamically adjusting applications written with the PETSc library. This model is based on historical performance information and data mining techniques. Finally, we demonstrate the validity of the proposed model through real experimentations.

Idioma original	Anglès nord-americà
Pàgines (de-a)	110-119
Nombre de pàgines	10
Revista	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
DOIs	https://doi.org/10.1007/978-3-540-85451-7_13
Estat de la publicació	Publicada - 2008

Accés al document

10.1007/978-3-540-85451-7_13

Altres arxius i enllaços

http://www.scopus.com/inward/record.url?scp=51849090611&partnerID=8YFLogxK

Com citar-ho

@article{8891b24c0ba34349a095bdf452487724,

title = "Performance model for parallel mathematical libraries based on historical knowledgebase",

abstract = "Scientific and mathematical parallel libraries offer a high level of abstraction to programmers. However, it is still difficult to select the proper parameters and algorithms to maximize the application performance. This work proposes a performance model for dynamically adjusting applications written with the PETSc library. This model is based on historical performance information and data mining techniques. Finally, we demonstrate the validity of the proposed model through real experimentations.",

keywords = "Dynamic mathematical model, Mathematical Performance, Performance Model, PETSc Performance",

author = "I. Salawdeh and E. C{\'e}sar and A. Morajko and T. Margalef and E. Luque",

year = "2008",

doi = "10.1007/978-3-540-85451-7_13",

language = "Ingl{\'e}s estadounidense",

pages = "110--119",

journal = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

issn = "0302-9743",

}

Performance model for parallel mathematical libraries based on historical knowledgebase. / Salawdeh, I.; César, E.; Morajko, A. et al.
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2008, pàg. 110-119.

Producció científica: Contribució a revista › Article › Recerca › Avaluat per experts

TY - JOUR

T1 - Performance model for parallel mathematical libraries based on historical knowledgebase

AU - Salawdeh, I.

AU - César, E.

AU - Morajko, A.

AU - Margalef, T.

AU - Luque, E.

PY - 2008

Y1 - 2008

N2 - Scientific and mathematical parallel libraries offer a high level of abstraction to programmers. However, it is still difficult to select the proper parameters and algorithms to maximize the application performance. This work proposes a performance model for dynamically adjusting applications written with the PETSc library. This model is based on historical performance information and data mining techniques. Finally, we demonstrate the validity of the proposed model through real experimentations.

AB - Scientific and mathematical parallel libraries offer a high level of abstraction to programmers. However, it is still difficult to select the proper parameters and algorithms to maximize the application performance. This work proposes a performance model for dynamically adjusting applications written with the PETSc library. This model is based on historical performance information and data mining techniques. Finally, we demonstrate the validity of the proposed model through real experimentations.

KW - Dynamic mathematical model

KW - Mathematical Performance

KW - Performance Model

KW - PETSc Performance

UR - http://www.scopus.com/inward/record.url?scp=51849090611&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-85451-7_13

DO - 10.1007/978-3-540-85451-7_13

M3 - Artículo

AN - SCOPUS:51849090611

SN - 0302-9743

SP - 110

EP - 119

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -

Performance model for parallel mathematical libraries based on historical knowledgebase

Resum

Accés al document

Altres arxius i enllaços

Fingerprint

Com citar-ho