A Portable and Heterogeneous LU Factorization on IRIS

Pedro Valero-Lara; Jungwon Kim; Jeffrey S. Vetter

doi:10.1007/978-3-031-31209-0_2

A Portable and Heterogeneous LU Factorization on IRIS

Pedro Valero-Lara, Jungwon Kim, Jeffrey S. Vetter

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

1 Scopus citations

Abstract

Here, the IRIS programming model is evaluated as a method to improve performance portability for heterogeneous systems that use LU matrix factorization. LU (lower-upper) factorization is considered one of the most important numerical linear algebra operations used in multiple high-performance computing and scientific applications. IRIS enables the separation of the algorithm’s definition from the tuning by using tasks + dependencies. This considerably reduces the effort required to achieve performance portability on heterogeneous systems. One IRIS code can use different settings depending on the underlying hardware features. Different configurations are evaluated on two different heterogeneous systems to achieve important speedups for the reference code with minimal changes to the source code.

Original language	English
Title of host publication	Euro-Par 2022
Subtitle of host publication	Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers
Editors	Jeremy Singer, Yehia Elkhatib, Dora Blanco Heras, Patrick Diehl, Nick Brown, Aleksandar Ilic
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	17-31
Number of pages	15
ISBN (Print)	9783031312083
DOIs	https://doi.org/10.1007/978-3-031-31209-0_2
State	Published - 2023
Event	28th International European Conference on Parallel and Distributed Computing , Euro-Par 2022 - Glasgow, United Kingdom Duration: Aug 22 2022 → Aug 26 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13835 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	28th International European Conference on Parallel and Distributed Computing , Euro-Par 2022
Country/Territory	United Kingdom
City	Glasgow
Period	08/22/22 → 08/26/22

Funding

J. Kim—Now at NVIDIA. Notice: This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid up, irrevocable, world-wide license to publish or reproduce the published form of the manuscript, or allow others to do so, for U.S. Government purposes. The DOE will provide public access to these results in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

Keywords

CPU
GPU
Heterogeneity
IRIS
LU factorization
Performance portability
Tasking

Access to Document

10.1007/978-3-031-31209-0_2

Cite this

Valero-Lara, P., Kim, J., & Vetter, J. S. (2023). A Portable and Heterogeneous LU Factorization on IRIS. In J. Singer, Y. Elkhatib, D. Blanco Heras, P. Diehl, N. Brown, & A. Ilic (Eds.), Euro-Par 2022: Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers (pp. 17-31). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13835 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-31209-0_2

Valero-Lara, Pedro ; Kim, Jungwon ; Vetter, Jeffrey S. / A Portable and Heterogeneous LU Factorization on IRIS. Euro-Par 2022: Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers. editor / Jeremy Singer ; Yehia Elkhatib ; Dora Blanco Heras ; Patrick Diehl ; Nick Brown ; Aleksandar Ilic. Springer Science and Business Media Deutschland GmbH, 2023. pp. 17-31 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4cadc2bb50b447a89c95bf5c4d7abb01,

title = "A Portable and Heterogeneous LU Factorization on IRIS",

abstract = "Here, the IRIS programming model is evaluated as a method to improve performance portability for heterogeneous systems that use LU matrix factorization. LU (lower-upper) factorization is considered one of the most important numerical linear algebra operations used in multiple high-performance computing and scientific applications. IRIS enables the separation of the algorithm{\textquoteright}s definition from the tuning by using tasks + dependencies. This considerably reduces the effort required to achieve performance portability on heterogeneous systems. One IRIS code can use different settings depending on the underlying hardware features. Different configurations are evaluated on two different heterogeneous systems to achieve important speedups for the reference code with minimal changes to the source code.",

keywords = "CPU, GPU, Heterogeneity, IRIS, LU factorization, Performance portability, Tasking",

author = "Pedro Valero-Lara and Jungwon Kim and Vetter, {Jeffrey S.}",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 28th International European Conference on Parallel and Distributed Computing , Euro-Par 2022 ; Conference date: 22-08-2022 Through 26-08-2022",

year = "2023",

doi = "10.1007/978-3-031-31209-0_2",

language = "English",

isbn = "9783031312083",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "17--31",

editor = "Jeremy Singer and Yehia Elkhatib and {Blanco Heras}, Dora and Patrick Diehl and Nick Brown and Aleksandar Ilic",

booktitle = "Euro-Par 2022",

}

Valero-Lara, P, Kim, J & Vetter, JS 2023, A Portable and Heterogeneous LU Factorization on IRIS. in J Singer, Y Elkhatib, D Blanco Heras, P Diehl, N Brown & A Ilic (eds), Euro-Par 2022: Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13835 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 17-31, 28th International European Conference on Parallel and Distributed Computing , Euro-Par 2022, Glasgow, United Kingdom, 08/22/22. https://doi.org/10.1007/978-3-031-31209-0_2

A Portable and Heterogeneous LU Factorization on IRIS. / Valero-Lara, Pedro; Kim, Jungwon; Vetter, Jeffrey S.
Euro-Par 2022: Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers. ed. / Jeremy Singer; Yehia Elkhatib; Dora Blanco Heras; Patrick Diehl; Nick Brown; Aleksandar Ilic. Springer Science and Business Media Deutschland GmbH, 2023. p. 17-31 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13835 LNCS).

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

TY - GEN

T1 - A Portable and Heterogeneous LU Factorization on IRIS

AU - Valero-Lara, Pedro

AU - Kim, Jungwon

AU - Vetter, Jeffrey S.

PY - 2023

Y1 - 2023

N2 - Here, the IRIS programming model is evaluated as a method to improve performance portability for heterogeneous systems that use LU matrix factorization. LU (lower-upper) factorization is considered one of the most important numerical linear algebra operations used in multiple high-performance computing and scientific applications. IRIS enables the separation of the algorithm’s definition from the tuning by using tasks + dependencies. This considerably reduces the effort required to achieve performance portability on heterogeneous systems. One IRIS code can use different settings depending on the underlying hardware features. Different configurations are evaluated on two different heterogeneous systems to achieve important speedups for the reference code with minimal changes to the source code.

AB - Here, the IRIS programming model is evaluated as a method to improve performance portability for heterogeneous systems that use LU matrix factorization. LU (lower-upper) factorization is considered one of the most important numerical linear algebra operations used in multiple high-performance computing and scientific applications. IRIS enables the separation of the algorithm’s definition from the tuning by using tasks + dependencies. This considerably reduces the effort required to achieve performance portability on heterogeneous systems. One IRIS code can use different settings depending on the underlying hardware features. Different configurations are evaluated on two different heterogeneous systems to achieve important speedups for the reference code with minimal changes to the source code.

KW - CPU

KW - GPU

KW - Heterogeneity

KW - IRIS

KW - LU factorization

KW - Performance portability

KW - Tasking

UR - http://www.scopus.com/inward/record.url?scp=85161420276&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-31209-0_2

DO - 10.1007/978-3-031-31209-0_2

M3 - Conference contribution

AN - SCOPUS:85161420276

SN - 9783031312083

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 17

EP - 31

BT - Euro-Par 2022

A2 - Singer, Jeremy

A2 - Elkhatib, Yehia

A2 - Blanco Heras, Dora

A2 - Diehl, Patrick

A2 - Brown, Nick

A2 - Ilic, Aleksandar

PB - Springer Science and Business Media Deutschland GmbH

T2 - 28th International European Conference on Parallel and Distributed Computing , Euro-Par 2022

Y2 - 22 August 2022 through 26 August 2022

ER -

Valero-Lara P, Kim J, Vetter JS. A Portable and Heterogeneous LU Factorization on IRIS. In Singer J, Elkhatib Y, Blanco Heras D, Diehl P, Brown N, Ilic A, editors, Euro-Par 2022: Parallel Processing Workshops - Euro-Par 2022 International Workshops, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2023. p. 17-31. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-31209-0_2