Sequential Task Flow Runtime Model Improvements and Limitations

Yu Pei; George Bosilca; Jack Dongarra

doi:10.1109/ROSS56639.2022.00009

Sequential Task Flow Runtime Model Improvements and Limitations

Yu Pei, George Bosilca, Jack Dongarra

Computer Science and Mathematics Div

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

1 Scopus citations

Abstract

The sequential task flow (STF) model is the main-stream approach for interacting with task-based runtime systems, with StarPU and the Dynamic task discovery (DTD) in PaRSEC being two implementations of this model. Compared with other approaches of submitting tasks into a runtime system, STF has interesting advantages centered around an easy-to-use API, that allows users to expressed algorithms as a sequence of tasks (much like in OpenMP), while allowing the runtime to automatically identify and analyze the task dependencies and scheduling. In this paper, we focus on the DTD interface in PaRSEC, highlight some of its lesser known limitations and implemented two optimization techniques for DTD: support for user level graph trimming, and a new API for broadcast read-only data to remote tasks. We then analyze the benefits and limitations of these optimizations with benchmarks as well as on two common matrix factorization kernels Cholesky and QR, on two different systems Shaheen II from KAUST and Fugaku from RIKEN. We point out some potential for further improvements, and provided valuable insights into the strength and weakness of STF model. hoping to guide the future developments of task-based runtime systems.

Original language	English
Title of host publication	Proceedings of ROSS 2022
Subtitle of host publication	International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	18-25
Number of pages	8
ISBN (Electronic)	9781665475662
DOIs	https://doi.org/10.1109/ROSS56639.2022.00009
State	Published - 2022
Event	12th IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers, ROSS 2022 - Dallas, United States Duration: Nov 13 2022 → Nov 18 2022

Publication series

Name	Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis

Conference

Conference	12th IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers, ROSS 2022
Country/Territory	United States
City	Dallas
Period	11/13/22 → 11/18/22

Funding

For computer time, this research used the resources of the Supercomputing Laboratory (KSL) Shaheen II at King Abdullah University of Science & Technology (KAUST) in Thuwal Saudi Arabia and the supercomputer Fugaku provided by RIKEN.

Keywords

Dynamic Runtime Systems
High Performance Computing
Numerical Linear Algebra

Access to Document

10.1109/ROSS56639.2022.00009

Cite this

Pei, Y., Bosilca, G., & Dongarra, J. (2022). Sequential Task Flow Runtime Model Improvements and Limitations. In Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis (pp. 18-25). (Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ROSS56639.2022.00009

Pei, Yu ; Bosilca, George ; Dongarra, Jack. / Sequential Task Flow Runtime Model Improvements and Limitations. Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 18-25 (Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis).

@inproceedings{80e3e23afa814b14a34d50739615cbd1,

title = "Sequential Task Flow Runtime Model Improvements and Limitations",

abstract = "The sequential task flow (STF) model is the main-stream approach for interacting with task-based runtime systems, with StarPU and the Dynamic task discovery (DTD) in PaRSEC being two implementations of this model. Compared with other approaches of submitting tasks into a runtime system, STF has interesting advantages centered around an easy-to-use API, that allows users to expressed algorithms as a sequence of tasks (much like in OpenMP), while allowing the runtime to automatically identify and analyze the task dependencies and scheduling. In this paper, we focus on the DTD interface in PaRSEC, highlight some of its lesser known limitations and implemented two optimization techniques for DTD: support for user level graph trimming, and a new API for broadcast read-only data to remote tasks. We then analyze the benefits and limitations of these optimizations with benchmarks as well as on two common matrix factorization kernels Cholesky and QR, on two different systems Shaheen II from KAUST and Fugaku from RIKEN. We point out some potential for further improvements, and provided valuable insights into the strength and weakness of STF model. hoping to guide the future developments of task-based runtime systems.",

keywords = "Dynamic Runtime Systems, High Performance Computing, Numerical Linear Algebra",

author = "Yu Pei and George Bosilca and Jack Dongarra",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 12th IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers, ROSS 2022 ; Conference date: 13-11-2022 Through 18-11-2022",

year = "2022",

doi = "10.1109/ROSS56639.2022.00009",

language = "English",

series = "Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis",

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

pages = "18--25",

booktitle = "Proceedings of ROSS 2022",

}

Pei, Y, Bosilca, G & Dongarra, J 2022, Sequential Task Flow Runtime Model Improvements and Limitations. in Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis. Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis, Institute of Electrical and Electronics Engineers Inc., pp. 18-25, 12th IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers, ROSS 2022, Dallas, United States, 11/13/22. https://doi.org/10.1109/ROSS56639.2022.00009

Sequential Task Flow Runtime Model Improvements and Limitations. / Pei, Yu; Bosilca, George; Dongarra, Jack.
Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2022. p. 18-25 (Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: 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 - Sequential Task Flow Runtime Model Improvements and Limitations

AU - Pei, Yu

AU - Bosilca, George

AU - Dongarra, Jack

PY - 2022

Y1 - 2022

N2 - The sequential task flow (STF) model is the main-stream approach for interacting with task-based runtime systems, with StarPU and the Dynamic task discovery (DTD) in PaRSEC being two implementations of this model. Compared with other approaches of submitting tasks into a runtime system, STF has interesting advantages centered around an easy-to-use API, that allows users to expressed algorithms as a sequence of tasks (much like in OpenMP), while allowing the runtime to automatically identify and analyze the task dependencies and scheduling. In this paper, we focus on the DTD interface in PaRSEC, highlight some of its lesser known limitations and implemented two optimization techniques for DTD: support for user level graph trimming, and a new API for broadcast read-only data to remote tasks. We then analyze the benefits and limitations of these optimizations with benchmarks as well as on two common matrix factorization kernels Cholesky and QR, on two different systems Shaheen II from KAUST and Fugaku from RIKEN. We point out some potential for further improvements, and provided valuable insights into the strength and weakness of STF model. hoping to guide the future developments of task-based runtime systems.

AB - The sequential task flow (STF) model is the main-stream approach for interacting with task-based runtime systems, with StarPU and the Dynamic task discovery (DTD) in PaRSEC being two implementations of this model. Compared with other approaches of submitting tasks into a runtime system, STF has interesting advantages centered around an easy-to-use API, that allows users to expressed algorithms as a sequence of tasks (much like in OpenMP), while allowing the runtime to automatically identify and analyze the task dependencies and scheduling. In this paper, we focus on the DTD interface in PaRSEC, highlight some of its lesser known limitations and implemented two optimization techniques for DTD: support for user level graph trimming, and a new API for broadcast read-only data to remote tasks. We then analyze the benefits and limitations of these optimizations with benchmarks as well as on two common matrix factorization kernels Cholesky and QR, on two different systems Shaheen II from KAUST and Fugaku from RIKEN. We point out some potential for further improvements, and provided valuable insights into the strength and weakness of STF model. hoping to guide the future developments of task-based runtime systems.

KW - Dynamic Runtime Systems

KW - High Performance Computing

KW - Numerical Linear Algebra

UR - http://www.scopus.com/inward/record.url?scp=85147797826&partnerID=8YFLogxK

U2 - 10.1109/ROSS56639.2022.00009

DO - 10.1109/ROSS56639.2022.00009

M3 - Conference contribution

AN - SCOPUS:85147797826

T3 - Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis

SP - 18

EP - 25

BT - Proceedings of ROSS 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 12th IEEE/ACM International Workshop on Runtime and Operating Systems for Supercomputers, ROSS 2022

Y2 - 13 November 2022 through 18 November 2022

ER -

Pei Y, Bosilca G, Dongarra J. Sequential Task Flow Runtime Model Improvements and Limitations. In Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc. 2022. p. 18-25. (Proceedings of ROSS 2022: International Workshop on Runtime and Operating Systems for Supercomputers, Held in conjunction with SC 2022: The International Conference for High Performance Computing, Networking, Storage and Analysis). doi: 10.1109/ROSS56639.2022.00009

Sequential Task Flow Runtime Model Improvements and Limitations

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this