GPAW optimized for Blue Gene/P using hybrid programming

Mads Ruben Burgdorff Kristensen; Hans Henrik Happe; Brian Vinter

doi:10.1109/IPDPS.2009.5160936

GPAW optimized for Blue Gene/P using hybrid programming

Mads Ruben Burgdorff Kristensen, Hans Henrik Happe, Brian Vinter

Datalogisk Institut

5 Citationer (Scopus)

2746 Downloads (Pure)

Abstract

In this work we present optimizations of a Grid-based projector-augmented wave method software, GPAW for the Blue Gene/P architecture. The improvements are achieved by exploring the advantage of shared and distributed memory programming also known as hybrid programming. The work focuses on optimizing a very time consuming operation in GPAW, the finite-different stencil operation, and different hybrid programming approaches are evaluated. The work succeeds in demonstrating a hybrid programming model which is clearly beneficial compared to the original flat programming model. In total an improvement of 1.94 compared to the original implementation is obtained. The results we demonstrate here are reasonably general and may be applied to other finite difference codes.

Originalsprog	Engelsk
Titel	Proceedings of the 2009 IEEE International Symposium on Parallel & Distributed Processing
Antal sider	6
Forlag	IEEE
Publikationsdato	2009
Sider	1-6
ISBN (Elektronisk)	978-1-4244-3751-1
DOI	https://doi.org/10.1109/IPDPS.2009.5160936
Status	Udgivet - 2009
Begivenhed	International Parallel and Distributed Processing Symposium (IPDPS 2009) - Rom, Italien Varighed: 23 maj 2009 → 29 maj 2009 Konferencens nummer: 23

Konference

Konference	International Parallel and Distributed Processing Symposium (IPDPS 2009)
Nummer	23
Land/Område	Italien
By	Rom
Periode	23/05/2009 → 29/05/2009

Emneord

Det Natur- og Biovidenskabelige Fakultet

Adgang til dokumentet

10.1109/IPDPS.2009.5160936

MainIEEEForlagets udgivne version, 136 KB

Citationsformater

@inproceedings{97c920f0fabc11de825d000ea68e967b,

title = "GPAW optimized for Blue Gene/P using hybrid programming",

abstract = "In this work we present optimizations of a Grid-based projector-augmented wave method software, GPAW for the Blue Gene/P architecture. The improvements are achieved by exploring the advantage of shared and distributed memory programming also known as hybrid programming. The work focuses on optimizing a very time consuming operation in GPAW, the finite-different stencil operation, and different hybrid programming approaches are evaluated. The work succeeds in demonstrating a hybrid programming model which is clearly beneficial compared to the original flat programming model. In total an improvement of 1.94 compared to the original implementation is obtained. The results we demonstrate here are reasonably general and may be applied to other finite difference codes.",

keywords = "Faculty of Science, HPC, Hybrid parallel programming, Parallel framework, GPAW",

author = "Kristensen, {Mads Ruben Burgdorff} and Happe, {Hans Henrik} and Brian Vinter",

year = "2009",

doi = "10.1109/IPDPS.2009.5160936",

language = "English",

pages = "1--6",

booktitle = "Proceedings of the 2009 IEEE International Symposium on Parallel & Distributed Processing",

publisher = "IEEE",

note = "International Parallel and Distributed Processing Symposium (IPDPS 2009) ; Conference date: 23-05-2009 Through 29-05-2009",

}

TY - GEN

T1 - GPAW optimized for Blue Gene/P using hybrid programming

AU - Kristensen, Mads Ruben Burgdorff

AU - Happe, Hans Henrik

AU - Vinter, Brian

N1 - Conference code: 23

PY - 2009

Y1 - 2009

N2 - In this work we present optimizations of a Grid-based projector-augmented wave method software, GPAW for the Blue Gene/P architecture. The improvements are achieved by exploring the advantage of shared and distributed memory programming also known as hybrid programming. The work focuses on optimizing a very time consuming operation in GPAW, the finite-different stencil operation, and different hybrid programming approaches are evaluated. The work succeeds in demonstrating a hybrid programming model which is clearly beneficial compared to the original flat programming model. In total an improvement of 1.94 compared to the original implementation is obtained. The results we demonstrate here are reasonably general and may be applied to other finite difference codes.

AB - In this work we present optimizations of a Grid-based projector-augmented wave method software, GPAW for the Blue Gene/P architecture. The improvements are achieved by exploring the advantage of shared and distributed memory programming also known as hybrid programming. The work focuses on optimizing a very time consuming operation in GPAW, the finite-different stencil operation, and different hybrid programming approaches are evaluated. The work succeeds in demonstrating a hybrid programming model which is clearly beneficial compared to the original flat programming model. In total an improvement of 1.94 compared to the original implementation is obtained. The results we demonstrate here are reasonably general and may be applied to other finite difference codes.

KW - Faculty of Science

KW - HPC

KW - Hybrid parallel programming

KW - Parallel framework

KW - GPAW

U2 - 10.1109/IPDPS.2009.5160936

DO - 10.1109/IPDPS.2009.5160936

M3 - Article in proceedings

SP - 1

EP - 6

BT - Proceedings of the 2009 IEEE International Symposium on Parallel & Distributed Processing

PB - IEEE

T2 - International Parallel and Distributed Processing Symposium (IPDPS 2009)

Y2 - 23 May 2009 through 29 May 2009

ER -

GPAW optimized for Blue Gene/P using hybrid programming

Abstract

Konference

Emneord

Adgang til dokumentet

Fingeraftryk

Citationsformater