A locality-aware dynamic thread scheduler for GPGPUs

Yu Hao Huang; Ying Yu Tseng; Hsien Kai Kuo; Ta Kan Yen; Bo Cheng Charles Lai

doi:10.1109/PDCAT.2013.46

A locality-aware dynamic thread scheduler for GPGPUs

Yu Hao Huang^*, Ying Yu Tseng, Hsien Kai Kuo, Ta Kan Yen, Bo Cheng Charles Lai

^*Corresponding author for this work

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

1 Scopus citations

Abstract

Modern GPGPUs implement on-chip shared cache to better exploit the data reuse of various general purpose applications. Given the massive amount of concurrent threads in a GPGPU, striking the balance between Data Locality and Load Balance has become a critical design concern. To achieve the best performance, the trade-off between these two factors needs to be performed concurrently. This paper proposes a dynamic thread scheduler which co-optimizes both the data locality and load balance on a GPGPU. The proposed approach is evaluated using three applications with various input datasets. The results show that the proposed approach reduces the overall execution cycles by up to 16% when compared with other approaches concerning only one objective.

Original language	English
Title of host publication	Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings
Editors	Shi-Jinn Horng
Publisher	IEEE Computer Society
Pages	254-258
Number of pages	5
ISBN (Electronic)	9781479924189
DOIs	https://doi.org/10.1109/PDCAT.2013.46
State	Published - 18 Sep 2014
Event	14th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2013 - Taipei, Taiwan Duration: 16 Dec 2013 → 18 Dec 2013

Publication series

Name	Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings

Conference

Conference	14th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2013
Country/Territory	Taiwan
City	Taipei
Period	16/12/13 → 18/12/13

Keywords

Data Locality
GPU
parallel computing

Access to Document

10.1109/PDCAT.2013.46

Cite this

Huang, Y. H., Tseng, Y. Y., Kuo, H. K., Yen, T. K., & Lai, B. C. C. (2014). A locality-aware dynamic thread scheduler for GPGPUs. In S-J. Horng (Ed.), Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings (pp. 254-258). [6904263] (Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings). IEEE Computer Society. https://doi.org/10.1109/PDCAT.2013.46

@inproceedings{a658bfb6c2ef4578b47a5aa37abc71f6,

title = "A locality-aware dynamic thread scheduler for GPGPUs",

abstract = "Modern GPGPUs implement on-chip shared cache to better exploit the data reuse of various general purpose applications. Given the massive amount of concurrent threads in a GPGPU, striking the balance between Data Locality and Load Balance has become a critical design concern. To achieve the best performance, the trade-off between these two factors needs to be performed concurrently. This paper proposes a dynamic thread scheduler which co-optimizes both the data locality and load balance on a GPGPU. The proposed approach is evaluated using three applications with various input datasets. The results show that the proposed approach reduces the overall execution cycles by up to 16% when compared with other approaches concerning only one objective.",

keywords = "Data Locality, GPU, parallel computing",

author = "Huang, {Yu Hao} and Tseng, {Ying Yu} and Kuo, {Hsien Kai} and Yen, {Ta Kan} and Lai, {Bo Cheng Charles}",

year = "2014",

month = sep,

day = "18",

doi = "10.1109/PDCAT.2013.46",

language = "English",

series = "Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings",

publisher = "IEEE Computer Society",

pages = "254--258",

editor = "Shi-Jinn Horng",

booktitle = "Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings",

address = "United States",

note = "null ; Conference date: 16-12-2013 Through 18-12-2013",

}

Huang, YH, Tseng, YY, Kuo, HK, Yen, TK & Lai, BCC 2014, A locality-aware dynamic thread scheduler for GPGPUs. in S-J Horng (ed.), Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings., 6904263, Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings, IEEE Computer Society, pp. 254-258, 14th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2013, Taipei, Taiwan, 16/12/13. https://doi.org/10.1109/PDCAT.2013.46

A locality-aware dynamic thread scheduler for GPGPUs. / Huang, Yu Hao; Tseng, Ying Yu; Kuo, Hsien Kai et al.

Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings. ed. / Shi-Jinn Horng. IEEE Computer Society, 2014. p. 254-258 6904263 (Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings).

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

TY - GEN

T1 - A locality-aware dynamic thread scheduler for GPGPUs

AU - Huang, Yu Hao

AU - Tseng, Ying Yu

AU - Kuo, Hsien Kai

AU - Yen, Ta Kan

AU - Lai, Bo Cheng Charles

PY - 2014/9/18

Y1 - 2014/9/18

N2 - Modern GPGPUs implement on-chip shared cache to better exploit the data reuse of various general purpose applications. Given the massive amount of concurrent threads in a GPGPU, striking the balance between Data Locality and Load Balance has become a critical design concern. To achieve the best performance, the trade-off between these two factors needs to be performed concurrently. This paper proposes a dynamic thread scheduler which co-optimizes both the data locality and load balance on a GPGPU. The proposed approach is evaluated using three applications with various input datasets. The results show that the proposed approach reduces the overall execution cycles by up to 16% when compared with other approaches concerning only one objective.

AB - Modern GPGPUs implement on-chip shared cache to better exploit the data reuse of various general purpose applications. Given the massive amount of concurrent threads in a GPGPU, striking the balance between Data Locality and Load Balance has become a critical design concern. To achieve the best performance, the trade-off between these two factors needs to be performed concurrently. This paper proposes a dynamic thread scheduler which co-optimizes both the data locality and load balance on a GPGPU. The proposed approach is evaluated using three applications with various input datasets. The results show that the proposed approach reduces the overall execution cycles by up to 16% when compared with other approaches concerning only one objective.

KW - Data Locality

KW - GPU

KW - parallel computing

UR - http://www.scopus.com/inward/record.url?scp=84907979673&partnerID=8YFLogxK

U2 - 10.1109/PDCAT.2013.46

DO - 10.1109/PDCAT.2013.46

M3 - Conference contribution

AN - SCOPUS:84907979673

T3 - Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings

SP - 254

EP - 258

BT - Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings

A2 - Horng, Shi-Jinn

PB - IEEE Computer Society

Y2 - 16 December 2013 through 18 December 2013

ER -

Huang YH, Tseng YY, Kuo HK, Yen TK, Lai BCC. A locality-aware dynamic thread scheduler for GPGPUs. In Horng S-J, editor, Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings. IEEE Computer Society. 2014. p. 254-258. 6904263. (Parallel and Distributed Computing, Applications and Technologies, PDCAT Proceedings). doi: 10.1109/PDCAT.2013.46

A locality-aware dynamic thread scheduler for GPGPUs

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this