Load balanced heterogeneous parallelism for finite difference problems on image denoising.

Ruben Laso; José Carlos Cabaleiro; Francisco F. Rivera; M. Carmen Muñiz; José A. Álvarez-Dios

doi:10.1002/cmm4.1089

Load balanced heterogeneous parallelism for finite difference problems on image denoising.

Ruben Laso
, José Carlos Cabaleiro
, Francisco F. Rivera
, M. Carmen Muñiz
, José A. Álvarez-Dios

Forschungsgruppe Scientific Computing

Veröffentlichungen: Beitrag in Fachzeitschrift › Artikel › Peer Reviewed

Abstract

In this work, we introduce a heterogeneous scheme for computing iterative (or time-step) methods based on finite differences, using an image denoising problem as case study. The idea of this proposal is to dynamically split the domain of the problem into smaller regions based on the CPU and GPU performance, balancing the workload between them. Results show that this approach improves the execution times compared with only use GPU, which is typically faster than CPU in this kind of problems. In our experiments, performance improvements go from 3%, in scenarios where CPU can only handle a little portion of workload, to more than 30%, when CPU can assume more work.

Originalsprache	Englisch
Aufsatznummer	e1089
Fachzeitschrift	Computational and mathematical methods
Jahrgang	3
Ausgabenummer	3
DOIs	https://doi.org/10.1002/cmm4.1089
Publikationsstatus	Veröffentlicht - Mai 2021

Fördermittel

This work has received financial support from the Ministerio de Economía, Industria y Competitividad within the project TIN2016‐76373‐P and network CAPAP‐H. It was also funded by the Consellería de Cultura, Educación e Ordenación Universitaria of Xunta de Galicia (accr. 2016‐2019, ED431G/08 and reference competitive group 2019‐2021, ED431C 2018/19). information Conseller?a de Cultura, Educaci?n e Ordenaci?n Universitaria, Xunta de Galicia, 2019-2021, ED431C 2018/19, accr.2016-2019, ED431G/08; Ministerio de Econom?a, Industria y Competitividad, Gobierno de Espa?a, TIN2016-76373-P and CAPAP-H networkThis work has received financial support from the Ministerio de Econom?a, Industria y Competitividad within the project TIN2016-76373-P and network CAPAP-H. It was also funded by the Conseller?a de Cultura, Educaci?n e Ordenaci?n Universitaria of Xunta de Galicia (accr. 2016-2019, ED431G/08 and reference competitive group 2019-2021, ED431C 2018/19).

ÖFOS 2012

102023 Supercomputing

Zugriff auf Dokument

10.1002/cmm4.1089

https://dblp.org/db/journals/cmm/cmm3.html#RodriguezCRMA21

Andere Dateien und Links

Verknüpfung zur Publikation in Scopus

Zitationsweisen

@article{fb10f5bf94944c16b2e31c93d4a0382b,

title = "Load balanced heterogeneous parallelism for finite difference problems on image denoising.",

abstract = "In this work, we introduce a heterogeneous scheme for computing iterative (or time-step) methods based on finite differences, using an image denoising problem as case study. The idea of this proposal is to dynamically split the domain of the problem into smaller regions based on the CPU and GPU performance, balancing the workload between them. Results show that this approach improves the execution times compared with only use GPU, which is typically faster than CPU in this kind of problems. In our experiments, performance improvements go from 3\%, in scenarios where CPU can only handle a little portion of workload, to more than 30\%, when CPU can assume more work.",

keywords = "CUDA, image processing, OpenMP, finite differences, heterogeneous parallelism",

author = "Ruben Laso and Cabaleiro, \{Jos{\'e} Carlos\} and Rivera, \{Francisco F.\} and Mu{\~n}iz, \{M. Carmen\} and {\'A}lvarez-Dios, \{Jos{\'e} A.\}",

note = "Publisher Copyright: {\textcopyright} 2020 John Wiley \& Sons, Ltd.",

year = "2021",

month = may,

doi = "10.1002/cmm4.1089",

language = "English",

volume = "3",

journal = "Computational and mathematical methods",

publisher = "John Wiley \& Sons, Ltd.",

number = "3",

}

TY - JOUR

T1 - Load balanced heterogeneous parallelism for finite difference problems on image denoising.

AU - Laso, Ruben

AU - Cabaleiro, José Carlos

AU - Rivera, Francisco F.

AU - Muñiz, M. Carmen

AU - Álvarez-Dios, José A.

PY - 2021/5

Y1 - 2021/5

N2 - In this work, we introduce a heterogeneous scheme for computing iterative (or time-step) methods based on finite differences, using an image denoising problem as case study. The idea of this proposal is to dynamically split the domain of the problem into smaller regions based on the CPU and GPU performance, balancing the workload between them. Results show that this approach improves the execution times compared with only use GPU, which is typically faster than CPU in this kind of problems. In our experiments, performance improvements go from 3%, in scenarios where CPU can only handle a little portion of workload, to more than 30%, when CPU can assume more work.

AB - In this work, we introduce a heterogeneous scheme for computing iterative (or time-step) methods based on finite differences, using an image denoising problem as case study. The idea of this proposal is to dynamically split the domain of the problem into smaller regions based on the CPU and GPU performance, balancing the workload between them. Results show that this approach improves the execution times compared with only use GPU, which is typically faster than CPU in this kind of problems. In our experiments, performance improvements go from 3%, in scenarios where CPU can only handle a little portion of workload, to more than 30%, when CPU can assume more work.

KW - CUDA

KW - image processing

KW - OpenMP

KW - finite differences

KW - heterogeneous parallelism

UR - https://www.scopus.com/pages/publications/85106259793

U2 - 10.1002/cmm4.1089

DO - 10.1002/cmm4.1089

M3 - Article

VL - 3

JO - Computational and mathematical methods

JF - Computational and mathematical methods

IS - 3

M1 - e1089

ER -

Load balanced heterogeneous parallelism for finite difference problems on image denoising.

Abstract

Fördermittel

ÖFOS 2012

Zugriff auf Dokument

Andere Dateien und Links

Fingerprint

Zitationsweisen