AMA: Asynchronous management of accelerators for task-based programming models

Judit Planas; Rosa M. Badia; Eduard Ayguade; Jesus Labarta

doi:10.1016/j.procs.2015.05.212

AMA: Asynchronous management of accelerators for task-based programming models

Judit Planas, Rosa M. Badia, Eduard Ayguade, Jesus Labarta

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

6 Scopus citations

Abstract

Computational science has benefited in the last years from emerging accelerators that increase the performance of scientific simulations, but using these devices hinders the programming task. This paper presents AMA: a set of optimization techniques to efficiently manage multiaccelerator systems. AMA maximizes the overlap of computation and communication in a blocking-free way. Then, we can use such spare time to do other work while waiting for device operations. Implemented on top of a task-based framework, the experimental evaluation of AMA on a quad-GPU node shows that we reach the performance of a hand-tuned native CUDA code, with the advantage of fully hiding the device management. In addition, we obtain up to more than 2x performance speed-up with respect to the original framework implementation.

Original language	English (US)
Title of host publication	Procedia Computer Science
Publisher	Elsevier BV
Pages	130-139
Number of pages	10
DOIs	https://doi.org/10.1016/j.procs.2015.05.212
State	Published - Jun 1 2015
Externally published	Yes

Bibliographical note

KAUST Repository Item: Exported on 2022-06-24
Acknowledgements: European Commission (HiPEAC-3 Network of Excellence, FP7-ICT 287759), Intel-BSC Exas-cale Lab and IBM/BSC Exascale Initiative collaboration, Spanish Ministry of Education (FPU), Computación de Altas Prestaciones VI (TIN2012-34557), Generalitat de Catalunya (2014-SGR-1051). We thank KAUST IT Research Computing for granting access to their machines.
This publication acknowledges KAUST support, but has no KAUST affiliated authors.

Access to Document

10.1016/j.procs.2015.05.212

Cite this

@inproceedings{41da99d9ea3d48e4b375f56112708f5e,

title = "AMA: Asynchronous management of accelerators for task-based programming models",

abstract = "Computational science has benefited in the last years from emerging accelerators that increase the performance of scientific simulations, but using these devices hinders the programming task. This paper presents AMA: a set of optimization techniques to efficiently manage multiaccelerator systems. AMA maximizes the overlap of computation and communication in a blocking-free way. Then, we can use such spare time to do other work while waiting for device operations. Implemented on top of a task-based framework, the experimental evaluation of AMA on a quad-GPU node shows that we reach the performance of a hand-tuned native CUDA code, with the advantage of fully hiding the device management. In addition, we obtain up to more than 2x performance speed-up with respect to the original framework implementation.",

author = "Judit Planas and Badia, {Rosa M.} and Eduard Ayguade and Jesus Labarta",

note = "KAUST Repository Item: Exported on 2022-06-24 Acknowledgements: European Commission (HiPEAC-3 Network of Excellence, FP7-ICT 287759), Intel-BSC Exas-cale Lab and IBM/BSC Exascale Initiative collaboration, Spanish Ministry of Education (FPU), Computaci{\'o}n de Altas Prestaciones VI (TIN2012-34557), Generalitat de Catalunya (2014-SGR-1051). We thank KAUST IT Research Computing for granting access to their machines. This publication acknowledges KAUST support, but has no KAUST affiliated authors.",

year = "2015",

month = jun,

day = "1",

doi = "10.1016/j.procs.2015.05.212",

language = "English (US)",

pages = "130--139",

booktitle = "Procedia Computer Science",

publisher = "Elsevier BV",

}

TY - GEN

T1 - AMA: Asynchronous management of accelerators for task-based programming models

AU - Planas, Judit

AU - Badia, Rosa M.

AU - Ayguade, Eduard

AU - Labarta, Jesus

N1 - KAUST Repository Item: Exported on 2022-06-24 Acknowledgements: European Commission (HiPEAC-3 Network of Excellence, FP7-ICT 287759), Intel-BSC Exas-cale Lab and IBM/BSC Exascale Initiative collaboration, Spanish Ministry of Education (FPU), Computación de Altas Prestaciones VI (TIN2012-34557), Generalitat de Catalunya (2014-SGR-1051). We thank KAUST IT Research Computing for granting access to their machines. This publication acknowledges KAUST support, but has no KAUST affiliated authors.

PY - 2015/6/1

Y1 - 2015/6/1

N2 - Computational science has benefited in the last years from emerging accelerators that increase the performance of scientific simulations, but using these devices hinders the programming task. This paper presents AMA: a set of optimization techniques to efficiently manage multiaccelerator systems. AMA maximizes the overlap of computation and communication in a blocking-free way. Then, we can use such spare time to do other work while waiting for device operations. Implemented on top of a task-based framework, the experimental evaluation of AMA on a quad-GPU node shows that we reach the performance of a hand-tuned native CUDA code, with the advantage of fully hiding the device management. In addition, we obtain up to more than 2x performance speed-up with respect to the original framework implementation.

AB - Computational science has benefited in the last years from emerging accelerators that increase the performance of scientific simulations, but using these devices hinders the programming task. This paper presents AMA: a set of optimization techniques to efficiently manage multiaccelerator systems. AMA maximizes the overlap of computation and communication in a blocking-free way. Then, we can use such spare time to do other work while waiting for device operations. Implemented on top of a task-based framework, the experimental evaluation of AMA on a quad-GPU node shows that we reach the performance of a hand-tuned native CUDA code, with the advantage of fully hiding the device management. In addition, we obtain up to more than 2x performance speed-up with respect to the original framework implementation.

UR - http://hdl.handle.net/10754/679324

UR - https://linkinghub.elsevier.com/retrieve/pii/S1877050915010200

UR - http://www.scopus.com/inward/record.url?scp=84939143628&partnerID=8YFLogxK

U2 - 10.1016/j.procs.2015.05.212

DO - 10.1016/j.procs.2015.05.212

M3 - Conference contribution

SP - 130

EP - 139

BT - Procedia Computer Science

PB - Elsevier BV

ER -

AMA: Asynchronous management of accelerators for task-based programming models

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this