Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC

Thomas Herault; Yves Robert; George Bosilca; Jack Dongarra

doi:10.1109/ScalA49573.2019.00010

Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC

Thomas Herault, Yves Robert, George Bosilca, Jack Dongarra

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

18 Scopus citations

Abstract

This paper introduces a generic and flexible matrix- matrix multiplication algorithm C = A × B for state-of-the-art computing platforms. Typically, these platforms are distributed- memory machines whose nodes are equipped with several ac- celerators. To the best of our knowledge, SLATE [9] is the only library that provides a publicly available implementation on such platforms, and it is currently limited to problem instances where the C matrix can entirely fit in the memory of the GPU accelera- tors. Our algorithm relies on the classical tile-based outer-product algorithm, but enhances it with several control dependencies to increase data re-use and to optimize communication flow from/to the accelerators within each node. The algorithm is written with the PARSEC runtime system, which allows for a fast and generic implementation, while achieving close-to-peak performance.

Original language	English
Title of host publication	Proceedings of ScalA 2019
Subtitle of host publication	10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	33-41
Number of pages	9
ISBN (Electronic)	9781728159898
DOIs	https://doi.org/10.1109/ScalA49573.2019.00010
State	Published - Nov 2019
Externally published	Yes
Event	10th IEEE/ACM Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, ScalA 2019 - Denver, United States Duration: Nov 18 2019 → …

Publication series

Name	Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis

Conference

Conference	10th IEEE/ACM Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, ScalA 2019
Country/Territory	United States
City	Denver
Period	11/18/19 → …

Funding

ACKNOWLEDGEMENT This research was supported by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration. It used resources of the Oak Ridge Leadership Computing Facility at ORNL, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725.

Keywords

Accelerator architectures
Linear Algebra
Run- time environment

Access to Document

10.1109/ScalA49573.2019.00010

Cite this

Herault, T., Robert, Y., Bosilca, G., & Dongarra, J. (2019). Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC. In Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis (pp. 33-41). Article 8948618 (Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ScalA49573.2019.00010

Herault, Thomas ; Robert, Yves ; Bosilca, George et al. / Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC. Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 33-41 (Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis).

@inproceedings{505535411a9c492ca8b7c9d8dfb76bc2,

title = "Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC",

abstract = "This paper introduces a generic and flexible matrix- matrix multiplication algorithm C = A × B for state-of-the-art computing platforms. Typically, these platforms are distributed- memory machines whose nodes are equipped with several ac- celerators. To the best of our knowledge, SLATE [9] is the only library that provides a publicly available implementation on such platforms, and it is currently limited to problem instances where the C matrix can entirely fit in the memory of the GPU accelera- tors. Our algorithm relies on the classical tile-based outer-product algorithm, but enhances it with several control dependencies to increase data re-use and to optimize communication flow from/to the accelerators within each node. The algorithm is written with the PARSEC runtime system, which allows for a fast and generic implementation, while achieving close-to-peak performance.",

keywords = "Accelerator architectures, Linear Algebra, Run- time environment",

author = "Thomas Herault and Yves Robert and George Bosilca and Jack Dongarra",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 10th IEEE/ACM Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, ScalA 2019 ; Conference date: 18-11-2019",

year = "2019",

month = nov,

doi = "10.1109/ScalA49573.2019.00010",

language = "English",

series = "Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "33--41",

booktitle = "Proceedings of ScalA 2019",

}

Herault, T, Robert, Y, Bosilca, G & Dongarra, J 2019, Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC. in Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis., 8948618, Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis, Institute of Electrical and Electronics Engineers Inc., pp. 33-41, 10th IEEE/ACM Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, ScalA 2019, Denver, United States, 11/18/19. https://doi.org/10.1109/ScalA49573.2019.00010

Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC. / Herault, Thomas; Robert, Yves; Bosilca, George et al.
Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2019. p. 33-41 8948618 (Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis).

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

TY - GEN

T1 - Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC

AU - Herault, Thomas

AU - Robert, Yves

AU - Bosilca, George

AU - Dongarra, Jack

PY - 2019/11

Y1 - 2019/11

N2 - This paper introduces a generic and flexible matrix- matrix multiplication algorithm C = A × B for state-of-the-art computing platforms. Typically, these platforms are distributed- memory machines whose nodes are equipped with several ac- celerators. To the best of our knowledge, SLATE [9] is the only library that provides a publicly available implementation on such platforms, and it is currently limited to problem instances where the C matrix can entirely fit in the memory of the GPU accelera- tors. Our algorithm relies on the classical tile-based outer-product algorithm, but enhances it with several control dependencies to increase data re-use and to optimize communication flow from/to the accelerators within each node. The algorithm is written with the PARSEC runtime system, which allows for a fast and generic implementation, while achieving close-to-peak performance.

AB - This paper introduces a generic and flexible matrix- matrix multiplication algorithm C = A × B for state-of-the-art computing platforms. Typically, these platforms are distributed- memory machines whose nodes are equipped with several ac- celerators. To the best of our knowledge, SLATE [9] is the only library that provides a publicly available implementation on such platforms, and it is currently limited to problem instances where the C matrix can entirely fit in the memory of the GPU accelera- tors. Our algorithm relies on the classical tile-based outer-product algorithm, but enhances it with several control dependencies to increase data re-use and to optimize communication flow from/to the accelerators within each node. The algorithm is written with the PARSEC runtime system, which allows for a fast and generic implementation, while achieving close-to-peak performance.

KW - Accelerator architectures

KW - Linear Algebra

KW - Run- time environment

UR - http://www.scopus.com/inward/record.url?scp=85078700382&partnerID=8YFLogxK

U2 - 10.1109/ScalA49573.2019.00010

DO - 10.1109/ScalA49573.2019.00010

M3 - Conference contribution

AN - SCOPUS:85078700382

T3 - Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis

SP - 33

EP - 41

BT - Proceedings of ScalA 2019

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 10th IEEE/ACM Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, ScalA 2019

Y2 - 18 November 2019

ER -

Herault T, Robert Y, Bosilca G, Dongarra J. Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC. In Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc. 2019. p. 33-41. 8948618. (Proceedings of ScalA 2019: 10th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems - Held in conjunction with SC 2019: The International Conference for High Performance Computing, Networking, Storage and Analysis). doi: 10.1109/ScalA49573.2019.00010

Generic Matrix Multiplication for Multi-GPU Accelerated Distributed-Memory Platforms over PaRSEC

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this