Parallelism orchestration using DoPE: The degree of parallelism executive

Arun Raman; Hanjun Kim; Taewook Oh; Jae W. Lee; David I. August

doi:10.1145/1993498.1993502

Parallelism orchestration using DoPE: The degree of parallelism executive

Arun Raman, Hanjun Kim, Taewook Oh, Jae W. Lee, David I. August

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

51 Scopus citations

Abstract

In writing parallel programs, programmers expose parallelism and optimize it to meet a particular performance goal on a single platform under an assumed set of workload characteristics. In the field, changing workload characteristics, new parallel platforms, and deployments with different performance goals make the programmer's development-time choices suboptimal. To address this problem, this paper presents the Degree of Parallelism Executive (DoPE), an API and run-time system that separates the concern of exposing parallelism from that of optimizing it. Using the DoPE API, the application developer expresses parallelism options. During program execution, DoPE's run-time system uses this information to dynamically optimize the parallelism options in response to the facts on the ground. We easily port several emerging parallel applications to DoPE's API and demonstrate the DoPE run-time system's effectiveness in dynamically optimizing the parallelism for a variety of performance goals.

Original language	English (US)
Title of host publication	PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation
Publisher	Association for Computing Machinery
Pages	26-37
Number of pages	12
ISBN (Print)	9781450306638
DOIs	https://doi.org/10.1145/1993498.1993502
State	Published - 2011
Event	32nd ACM Conference on Programming Language Design and Implementation, PLDI 2011 - San Jose, United States Duration: Jun 4 2011 → Jun 8 2011

Publication series

Name	Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)

Conference

Conference	32nd ACM Conference on Programming Language Design and Implementation, PLDI 2011
Country/Territory	United States
City	San Jose
Period	6/4/11 → 6/8/11

All Science Journal Classification (ASJC) codes

Software

Keywords

dynamic
loop nest
loop-level
nested
optimization
parallelism
parallelization
parametric
pipeline
run-time
scheduling
task

Access to Document

10.1145/1993498.1993502

Cite this

Raman, A., Kim, H., Oh, T., Lee, J. W., & August, D. I. (2011). Parallelism orchestration using DoPE: The degree of parallelism executive. In PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation (pp. 26-37). (Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)). Association for Computing Machinery. https://doi.org/10.1145/1993498.1993502

Raman, Arun ; Kim, Hanjun ; Oh, Taewook et al. / Parallelism orchestration using DoPE : The degree of parallelism executive. PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation. Association for Computing Machinery, 2011. pp. 26-37 (Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)).

@inproceedings{ad0285dc2b63416d87536592228eb63f,

title = "Parallelism orchestration using DoPE: The degree of parallelism executive",

abstract = "In writing parallel programs, programmers expose parallelism and optimize it to meet a particular performance goal on a single platform under an assumed set of workload characteristics. In the field, changing workload characteristics, new parallel platforms, and deployments with different performance goals make the programmer's development-time choices suboptimal. To address this problem, this paper presents the Degree of Parallelism Executive (DoPE), an API and run-time system that separates the concern of exposing parallelism from that of optimizing it. Using the DoPE API, the application developer expresses parallelism options. During program execution, DoPE's run-time system uses this information to dynamically optimize the parallelism options in response to the facts on the ground. We easily port several emerging parallel applications to DoPE's API and demonstrate the DoPE run-time system's effectiveness in dynamically optimizing the parallelism for a variety of performance goals.",

keywords = "dynamic, loop nest, loop-level, nested, optimization, parallelism, parallelization, parametric, pipeline, run-time, scheduling, task",

author = "Arun Raman and Hanjun Kim and Taewook Oh and Lee, \{Jae W.\} and August, \{David I.\}",

year = "2011",

doi = "10.1145/1993498.1993502",

language = "English (US)",

isbn = "9781450306638",

series = "Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)",

publisher = "Association for Computing Machinery",

pages = "26--37",

booktitle = "PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation",

note = "32nd ACM Conference on Programming Language Design and Implementation, PLDI 2011 ; Conference date: 04-06-2011 Through 08-06-2011",

}

Raman, A, Kim, H, Oh, T, Lee, JW & August, DI 2011, Parallelism orchestration using DoPE: The degree of parallelism executive. in PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation. Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), Association for Computing Machinery, pp. 26-37, 32nd ACM Conference on Programming Language Design and Implementation, PLDI 2011, San Jose, United States, 6/4/11. https://doi.org/10.1145/1993498.1993502

Parallelism orchestration using DoPE: The degree of parallelism executive. / Raman, Arun; Kim, Hanjun; Oh, Taewook et al.
PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation. Association for Computing Machinery, 2011. p. 26-37 (Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)).

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

TY - GEN

T1 - Parallelism orchestration using DoPE

T2 - 32nd ACM Conference on Programming Language Design and Implementation, PLDI 2011

AU - Raman, Arun

AU - Kim, Hanjun

AU - Oh, Taewook

AU - Lee, Jae W.

AU - August, David I.

PY - 2011

Y1 - 2011

N2 - In writing parallel programs, programmers expose parallelism and optimize it to meet a particular performance goal on a single platform under an assumed set of workload characteristics. In the field, changing workload characteristics, new parallel platforms, and deployments with different performance goals make the programmer's development-time choices suboptimal. To address this problem, this paper presents the Degree of Parallelism Executive (DoPE), an API and run-time system that separates the concern of exposing parallelism from that of optimizing it. Using the DoPE API, the application developer expresses parallelism options. During program execution, DoPE's run-time system uses this information to dynamically optimize the parallelism options in response to the facts on the ground. We easily port several emerging parallel applications to DoPE's API and demonstrate the DoPE run-time system's effectiveness in dynamically optimizing the parallelism for a variety of performance goals.

AB - In writing parallel programs, programmers expose parallelism and optimize it to meet a particular performance goal on a single platform under an assumed set of workload characteristics. In the field, changing workload characteristics, new parallel platforms, and deployments with different performance goals make the programmer's development-time choices suboptimal. To address this problem, this paper presents the Degree of Parallelism Executive (DoPE), an API and run-time system that separates the concern of exposing parallelism from that of optimizing it. Using the DoPE API, the application developer expresses parallelism options. During program execution, DoPE's run-time system uses this information to dynamically optimize the parallelism options in response to the facts on the ground. We easily port several emerging parallel applications to DoPE's API and demonstrate the DoPE run-time system's effectiveness in dynamically optimizing the parallelism for a variety of performance goals.

KW - dynamic

KW - loop nest

KW - loop-level

KW - nested

KW - optimization

KW - parallelism

KW - parallelization

KW - parametric

KW - pipeline

KW - run-time

KW - scheduling

KW - task

UR - https://www.scopus.com/pages/publications/79959909380

UR - https://www.scopus.com/inward/citedby.url?scp=79959909380&partnerID=8YFLogxK

U2 - 10.1145/1993498.1993502

DO - 10.1145/1993498.1993502

M3 - Conference contribution

AN - SCOPUS:79959909380

SN - 9781450306638

T3 - Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)

SP - 26

EP - 37

BT - PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation

PB - Association for Computing Machinery

Y2 - 4 June 2011 through 8 June 2011

ER -

Raman A, Kim H, Oh T, Lee JW, August DI. Parallelism orchestration using DoPE: The degree of parallelism executive. In PLDI'11 - Proceedings of the 2011 ACM Conference on Programming Language Design and Implementation. Association for Computing Machinery. 2011. p. 26-37. (Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI)). doi: 10.1145/1993498.1993502

Parallelism orchestration using DoPE: The degree of parallelism executive

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Keywords

Access to Document

Other files and links

Fingerprint

Cite this