SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling

Bo Wu; Guoyang Chen; Dong Li; Xipeng Shen; Jeffrey S. Vetter

doi:10.1145/2628071.2628130

SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling

Bo Wu, Guoyang Chen, Dong Li, Xipeng Shen, Jeffrey S. Vetter

Computer Science and Mathematics Div

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

1 Scopus citations

Abstract

To circumvent the limitation from the hardware scheduler on GPU, we create an SM-centric transformation technique. This technique enables complete control of the mapping between tasks and streaming multi-processors (SMs), and enables controlling the number of active thread blocks on each SM. Results show that our approach achieves better speedup than previous ones with kernel co-run cases.

Original language	English
Title of host publication	PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	497-498
Number of pages	2
ISBN (Print)	9781450328098
DOIs	https://doi.org/10.1145/2628071.2628130
State	Published - 2014
Event	23rd International Conference on Parallel Architectures and Compilation Techniques, PACT 2014 - Edmonton, AB, Canada Duration: Aug 24 2014 → Aug 27 2014

Publication series

Name	Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT
ISSN (Print)	1089-795X

Conference

Conference	23rd International Conference on Parallel Architectures and Compilation Techniques, PACT 2014
Country/Territory	Canada
City	Edmonton, AB
Period	08/24/14 → 08/27/14

Keywords

gpgpu
program co-run
scheduling

Access to Document

10.1145/2628071.2628130

Cite this

Wu, B., Chen, G., Li, D., Shen, X., & Vetter, J. S. (2014). SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling. In PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques (pp. 497-498). (Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/2628071.2628130

Wu, Bo ; Chen, Guoyang ; Li, Dong et al. / SM-centric transformation : Circumventing hardware restrictions for flexible GPU scheduling. PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques. Institute of Electrical and Electronics Engineers Inc., 2014. pp. 497-498 (Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT).

@inproceedings{4d04068011fd464fa6dd3ef8c47abd19,

title = "SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling",

abstract = "To circumvent the limitation from the hardware scheduler on GPU, we create an SM-centric transformation technique. This technique enables complete control of the mapping between tasks and streaming multi-processors (SMs), and enables controlling the number of active thread blocks on each SM. Results show that our approach achieves better speedup than previous ones with kernel co-run cases.",

keywords = "gpgpu, program co-run, scheduling",

author = "Bo Wu and Guoyang Chen and Dong Li and Xipeng Shen and Vetter, {Jeffrey S.}",

year = "2014",

doi = "10.1145/2628071.2628130",

language = "English",

isbn = "9781450328098",

series = "Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT",

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

pages = "497--498",

booktitle = "PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques",

note = "23rd International Conference on Parallel Architectures and Compilation Techniques, PACT 2014 ; Conference date: 24-08-2014 Through 27-08-2014",

}

Wu, B, Chen, G, Li, D, Shen, X & Vetter, JS 2014, SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling. in PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques. Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT, Institute of Electrical and Electronics Engineers Inc., pp. 497-498, 23rd International Conference on Parallel Architectures and Compilation Techniques, PACT 2014, Edmonton, AB, Canada, 08/24/14. https://doi.org/10.1145/2628071.2628130

SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling. / Wu, Bo; Chen, Guoyang; Li, Dong et al.
PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques. Institute of Electrical and Electronics Engineers Inc., 2014. p. 497-498 (Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT).

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

TY - GEN

T1 - SM-centric transformation

T2 - 23rd International Conference on Parallel Architectures and Compilation Techniques, PACT 2014

AU - Wu, Bo

AU - Chen, Guoyang

AU - Li, Dong

AU - Shen, Xipeng

AU - Vetter, Jeffrey S.

PY - 2014

Y1 - 2014

N2 - To circumvent the limitation from the hardware scheduler on GPU, we create an SM-centric transformation technique. This technique enables complete control of the mapping between tasks and streaming multi-processors (SMs), and enables controlling the number of active thread blocks on each SM. Results show that our approach achieves better speedup than previous ones with kernel co-run cases.

AB - To circumvent the limitation from the hardware scheduler on GPU, we create an SM-centric transformation technique. This technique enables complete control of the mapping between tasks and streaming multi-processors (SMs), and enables controlling the number of active thread blocks on each SM. Results show that our approach achieves better speedup than previous ones with kernel co-run cases.

KW - gpgpu

KW - program co-run

KW - scheduling

UR - http://www.scopus.com/inward/record.url?scp=84907083862&partnerID=8YFLogxK

U2 - 10.1145/2628071.2628130

DO - 10.1145/2628071.2628130

M3 - Conference contribution

AN - SCOPUS:84907083862

SN - 9781450328098

T3 - Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT

SP - 497

EP - 498

BT - PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 24 August 2014 through 27 August 2014

ER -

Wu B, Chen G, Li D, Shen X, Vetter JS. SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling. In PACT 2014 - Proceedings of the 23rd International Conference on Parallel Architectures and Compilation Techniques. Institute of Electrical and Electronics Engineers Inc. 2014. p. 497-498. (Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT). doi: 10.1145/2628071.2628130

SM-centric transformation: Circumventing hardware restrictions for flexible GPU scheduling

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this