Leveraging MPI's one-sided communication interface for shared-memory programming

Torsten Hoefler; James Dinan; Darius Buntinas; Pavan Balaji; Brian W. Barrett; Ron Brightwell; William Gropp; Vivek Kale; Rajeev Thakur

doi:10.1007/978-3-642-33518-1_18

Leveraging MPI's one-sided communication interface for shared-memory programming

Torsten Hoefler, James Dinan, Darius Buntinas, Pavan Balaji, Brian W. Barrett, Ron Brightwell, William Gropp, Vivek Kale, Rajeev Thakur

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

Abstract

Hybrid parallel programming with MPI for internode communication in conjunction with a shared-memory programming model to manage intranode parallelism has become a dominant approach to scalable parallel programming. While this model provides a great deal of flexibility and performance potential, it saddles programmers with the complexity of utilizing two parallel programming systems in the same application. We introduce an MPI-integrated shared-memory programming model that is incorporated into MPI through a small extension to the one-sided communication interface. We discuss the integration of this interface with the upcoming MPI 3.0 one-sided semantics and describe solutions for providing portable and efficient data sharing, atomic operations, and memory consistency. We describe an implementation of the new interface in the MPICH2 and Open MPI implementations and demonstrate an average performance improvement of 40% to the communication component of a five-point stencil solver.

Original language	English (US)
Title of host publication	Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings
Pages	132-141
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-33518-1_18
State	Published - 2012
Event	19th European MPI Users' Group Meeting on Recent Advances in the Message Passing Interface, EuroMPI 2012 - Vienna, Austria Duration: Sep 23 2012 → Sep 26 2012

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7490 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	19th European MPI Users' Group Meeting on Recent Advances in the Message Passing Interface, EuroMPI 2012
Country/Territory	Austria
City	Vienna
Period	9/23/12 → 9/26/12

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Online availability

10.1007/978-3-642-33518-1_18

Library availability

Discover UIUC Full Text

Cite this

Hoefler, T., Dinan, J., Buntinas, D., Balaji, P., Barrett, B. W., Brightwell, R., Gropp, W., Kale, V., & Thakur, R. (2012). Leveraging MPI's one-sided communication interface for shared-memory programming. In Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings (pp. 132-141). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7490 LNCS). https://doi.org/10.1007/978-3-642-33518-1_18

Leveraging MPI's one-sided communication interface for shared-memory programming. / Hoefler, Torsten; Dinan, James; Buntinas, Darius et al.
Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings. 2012. p. 132-141 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7490 LNCS).

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

Hoefler, T, Dinan, J, Buntinas, D, Balaji, P, Barrett, BW, Brightwell, R, Gropp, W, Kale, V & Thakur, R 2012, Leveraging MPI's one-sided communication interface for shared-memory programming. in Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7490 LNCS, pp. 132-141, 19th European MPI Users' Group Meeting on Recent Advances in the Message Passing Interface, EuroMPI 2012, Vienna, Austria, 9/23/12. https://doi.org/10.1007/978-3-642-33518-1_18

Hoefler T, Dinan J, Buntinas D, Balaji P, Barrett BW, Brightwell R et al. Leveraging MPI's one-sided communication interface for shared-memory programming. In Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings. 2012. p. 132-141. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-33518-1_18

Hoefler, Torsten ; Dinan, James ; Buntinas, Darius et al. / Leveraging MPI's one-sided communication interface for shared-memory programming. Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings. 2012. pp. 132-141 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{e70cfe3b2cc044c388d27c454bfbc505,

title = "Leveraging MPI's one-sided communication interface for shared-memory programming",

abstract = "Hybrid parallel programming with MPI for internode communication in conjunction with a shared-memory programming model to manage intranode parallelism has become a dominant approach to scalable parallel programming. While this model provides a great deal of flexibility and performance potential, it saddles programmers with the complexity of utilizing two parallel programming systems in the same application. We introduce an MPI-integrated shared-memory programming model that is incorporated into MPI through a small extension to the one-sided communication interface. We discuss the integration of this interface with the upcoming MPI 3.0 one-sided semantics and describe solutions for providing portable and efficient data sharing, atomic operations, and memory consistency. We describe an implementation of the new interface in the MPICH2 and Open MPI implementations and demonstrate an average performance improvement of 40% to the communication component of a five-point stencil solver.",

author = "Torsten Hoefler and James Dinan and Darius Buntinas and Pavan Balaji and Barrett, {Brian W.} and Ron Brightwell and William Gropp and Vivek Kale and Rajeev Thakur",

year = "2012",

doi = "10.1007/978-3-642-33518-1_18",

language = "English (US)",

isbn = "9783642335174",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "132--141",

booktitle = "Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings",

note = "19th European MPI Users' Group Meeting on Recent Advances in the Message Passing Interface, EuroMPI 2012 ; Conference date: 23-09-2012 Through 26-09-2012",

}

TY - GEN

T1 - Leveraging MPI's one-sided communication interface for shared-memory programming

AU - Hoefler, Torsten

AU - Dinan, James

AU - Buntinas, Darius

AU - Balaji, Pavan

AU - Barrett, Brian W.

AU - Brightwell, Ron

AU - Gropp, William

AU - Kale, Vivek

AU - Thakur, Rajeev

PY - 2012

Y1 - 2012

N2 - Hybrid parallel programming with MPI for internode communication in conjunction with a shared-memory programming model to manage intranode parallelism has become a dominant approach to scalable parallel programming. While this model provides a great deal of flexibility and performance potential, it saddles programmers with the complexity of utilizing two parallel programming systems in the same application. We introduce an MPI-integrated shared-memory programming model that is incorporated into MPI through a small extension to the one-sided communication interface. We discuss the integration of this interface with the upcoming MPI 3.0 one-sided semantics and describe solutions for providing portable and efficient data sharing, atomic operations, and memory consistency. We describe an implementation of the new interface in the MPICH2 and Open MPI implementations and demonstrate an average performance improvement of 40% to the communication component of a five-point stencil solver.

AB - Hybrid parallel programming with MPI for internode communication in conjunction with a shared-memory programming model to manage intranode parallelism has become a dominant approach to scalable parallel programming. While this model provides a great deal of flexibility and performance potential, it saddles programmers with the complexity of utilizing two parallel programming systems in the same application. We introduce an MPI-integrated shared-memory programming model that is incorporated into MPI through a small extension to the one-sided communication interface. We discuss the integration of this interface with the upcoming MPI 3.0 one-sided semantics and describe solutions for providing portable and efficient data sharing, atomic operations, and memory consistency. We describe an implementation of the new interface in the MPICH2 and Open MPI implementations and demonstrate an average performance improvement of 40% to the communication component of a five-point stencil solver.

UR - http://www.scopus.com/inward/record.url?scp=84867646537&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867646537&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-33518-1_18

DO - 10.1007/978-3-642-33518-1_18

M3 - Conference contribution

AN - SCOPUS:84867646537

SN - 9783642335174

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 132

EP - 141

BT - Recent Advances in the Message Passing Interface - 19th European MPI Users' Group Meeting, EuroMPI 2012, Proceedings

T2 - 19th European MPI Users' Group Meeting on Recent Advances in the Message Passing Interface, EuroMPI 2012

Y2 - 23 September 2012 through 26 September 2012

ER -

Leveraging MPI's one-sided communication interface for shared-memory programming

Abstract

Publication series

Other

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this