The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs

Yuki Tsujita, Toshio Endo, Katsuki Fujisawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

The Cholesky factorization is an important linear algebra kernel which is used in the semidefinite programming (SDP) problem. However, the large computation costs for Cholesky factorization of the Schur complement matrix (SCM) has been obstacles to solve large scale problems. This paper describes a brand-new version of the parallel SDP solver, SDPARA, which has been equipped with a Cholesky factorization implementation and demonstrated 1.7PFlops performance with over two million constraints by using 4,080 GPUs. The performance and scalability is even more improved by introducing a data-driven approach, rather than traditional synchronous approach. Also we point out that typical data-driven implementations have limitation in scalability, and demonstrate the efficiency of the proposed approach via experiments on TSUBAME2.5 supercomputer.

Original languageEnglish
Title of host publicationProceedings of ESPM2 2015
Subtitle of host publication1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis
PublisherAssociation for Computing Machinery, Inc
Pages38-45
Number of pages8
ISBN (Electronic)9781450339964
DOIs
Publication statusPublished - Nov 15 2015
Event1st International Workshop on Extreme Scale Programming Models and Middleware, ESPM2 2015 - Austin, United States
Duration: Nov 15 2015 → …

Publication series

NameProceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis

Other

Other1st International Workshop on Extreme Scale Programming Models and Middleware, ESPM2 2015
CountryUnited States
CityAustin
Period11/15/15 → …

Fingerprint

Factorization
Scalability
Parallel programming
Linear algebra
Supercomputers
Graphics processing unit
Costs
Experiments

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Software

Cite this

Tsujita, Y., Endo, T., & Fujisawa, K. (2015). The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs. In Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis (pp. 38-45). (Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis). Association for Computing Machinery, Inc. https://doi.org/10.1145/2832241.2832245

The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs. / Tsujita, Yuki; Endo, Toshio; Fujisawa, Katsuki.

Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis. Association for Computing Machinery, Inc, 2015. p. 38-45 (Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Tsujita, Y, Endo, T & Fujisawa, K 2015, The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs. in Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis. Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis, Association for Computing Machinery, Inc, pp. 38-45, 1st International Workshop on Extreme Scale Programming Models and Middleware, ESPM2 2015, Austin, United States, 11/15/15. https://doi.org/10.1145/2832241.2832245
Tsujita Y, Endo T, Fujisawa K. The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs. In Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis. Association for Computing Machinery, Inc. 2015. p. 38-45. (Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis). https://doi.org/10.1145/2832241.2832245
Tsujita, Yuki ; Endo, Toshio ; Fujisawa, Katsuki. / The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs. Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis. Association for Computing Machinery, Inc, 2015. pp. 38-45 (Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis).
@inproceedings{f5a551998c114c899477ca6093313a25,
title = "The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs",
abstract = "The Cholesky factorization is an important linear algebra kernel which is used in the semidefinite programming (SDP) problem. However, the large computation costs for Cholesky factorization of the Schur complement matrix (SCM) has been obstacles to solve large scale problems. This paper describes a brand-new version of the parallel SDP solver, SDPARA, which has been equipped with a Cholesky factorization implementation and demonstrated 1.7PFlops performance with over two million constraints by using 4,080 GPUs. The performance and scalability is even more improved by introducing a data-driven approach, rather than traditional synchronous approach. Also we point out that typical data-driven implementations have limitation in scalability, and demonstrate the efficiency of the proposed approach via experiments on TSUBAME2.5 supercomputer.",
author = "Yuki Tsujita and Toshio Endo and Katsuki Fujisawa",
year = "2015",
month = "11",
day = "15",
doi = "10.1145/2832241.2832245",
language = "English",
series = "Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis",
publisher = "Association for Computing Machinery, Inc",
pages = "38--45",
booktitle = "Proceedings of ESPM2 2015",

}

TY - GEN

T1 - The scalable petascale data-driven approach for the Cholesky factorization with multiple GPUs

AU - Tsujita, Yuki

AU - Endo, Toshio

AU - Fujisawa, Katsuki

PY - 2015/11/15

Y1 - 2015/11/15

N2 - The Cholesky factorization is an important linear algebra kernel which is used in the semidefinite programming (SDP) problem. However, the large computation costs for Cholesky factorization of the Schur complement matrix (SCM) has been obstacles to solve large scale problems. This paper describes a brand-new version of the parallel SDP solver, SDPARA, which has been equipped with a Cholesky factorization implementation and demonstrated 1.7PFlops performance with over two million constraints by using 4,080 GPUs. The performance and scalability is even more improved by introducing a data-driven approach, rather than traditional synchronous approach. Also we point out that typical data-driven implementations have limitation in scalability, and demonstrate the efficiency of the proposed approach via experiments on TSUBAME2.5 supercomputer.

AB - The Cholesky factorization is an important linear algebra kernel which is used in the semidefinite programming (SDP) problem. However, the large computation costs for Cholesky factorization of the Schur complement matrix (SCM) has been obstacles to solve large scale problems. This paper describes a brand-new version of the parallel SDP solver, SDPARA, which has been equipped with a Cholesky factorization implementation and demonstrated 1.7PFlops performance with over two million constraints by using 4,080 GPUs. The performance and scalability is even more improved by introducing a data-driven approach, rather than traditional synchronous approach. Also we point out that typical data-driven implementations have limitation in scalability, and demonstrate the efficiency of the proposed approach via experiments on TSUBAME2.5 supercomputer.

UR - http://www.scopus.com/inward/record.url?scp=85009107942&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85009107942&partnerID=8YFLogxK

U2 - 10.1145/2832241.2832245

DO - 10.1145/2832241.2832245

M3 - Conference contribution

AN - SCOPUS:85009107942

T3 - Proceedings of ESPM2 2015: 1st International Workshop on Extreme Scale Programming Models and Middleware - Held in conjunction with SC 2015: The International Conference for High Performance Computing, Networking, Storage and Analysis

SP - 38

EP - 45

BT - Proceedings of ESPM2 2015

PB - Association for Computing Machinery, Inc

ER -