Locality and topology aware intra-node communication among multicore CPUs

Teng Ma; George Bosilca; Aurelien Bouteiller; Jack J. Dongarra

doi:10.1007/978-3-642-15646-5_28

Locality and topology aware intra-node communication among multicore CPUs

Teng Ma, George Bosilca, Aurelien Bouteiller, Jack J. Dongarra

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

16 Scopus citations

Abstract

A major trend in HPC is the escalation toward manycore, where systems are composed of shared memory nodes featuring numerous processing units. Unfortunately, with scale comes complexity, here in the form of non-uniform memory accesses and cache hierarchies. For most HPC applications, harnessing the power of multicores is hindered by the topology oblivious tuning of the MPI library. In this paper, we propose a framework to tune every type of shared memory communications according to locality and topology. An implementation inside Open MPI is evaluated experimentally and demonstrates significant speedups compared to vanilla Open MPI and MPICH2.

Original language	English
Title of host publication	Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings
Pages	265-274
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-15646-5_28
State	Published - 2010
Externally published	Yes
Event	17th European MPI Users' Group Meeting, EuroMPI 2010 - Stuttgart, Germany Duration: Sep 12 2010 → Sep 15 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	6305 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th European MPI Users' Group Meeting, EuroMPI 2010
Country/Territory	Germany
City	Stuttgart
Period	09/12/10 → 09/15/10

Access to Document

10.1007/978-3-642-15646-5_28

Cite this

Ma, T., Bosilca, G., Bouteiller, A., & Dongarra, J. J. (2010). Locality and topology aware intra-node communication among multicore CPUs. In Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings (pp. 265-274). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6305 LNCS). https://doi.org/10.1007/978-3-642-15646-5_28

Ma, Teng ; Bosilca, George ; Bouteiller, Aurelien et al. / Locality and topology aware intra-node communication among multicore CPUs. Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. 2010. pp. 265-274 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b9de7697642c4c80b2586c7ea49c9ef7,

title = "Locality and topology aware intra-node communication among multicore CPUs",

abstract = "A major trend in HPC is the escalation toward manycore, where systems are composed of shared memory nodes featuring numerous processing units. Unfortunately, with scale comes complexity, here in the form of non-uniform memory accesses and cache hierarchies. For most HPC applications, harnessing the power of multicores is hindered by the topology oblivious tuning of the MPI library. In this paper, we propose a framework to tune every type of shared memory communications according to locality and topology. An implementation inside Open MPI is evaluated experimentally and demonstrates significant speedups compared to vanilla Open MPI and MPICH2.",

author = "Teng Ma and George Bosilca and Aurelien Bouteiller and Dongarra, {Jack J.}",

year = "2010",

doi = "10.1007/978-3-642-15646-5_28",

language = "English",

isbn = "3642156452",

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

pages = "265--274",

booktitle = "Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings",

note = "17th European MPI Users' Group Meeting, EuroMPI 2010 ; Conference date: 12-09-2010 Through 15-09-2010",

}

Ma, T, Bosilca, G, Bouteiller, A & Dongarra, JJ 2010, Locality and topology aware intra-node communication among multicore CPUs. in Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6305 LNCS, pp. 265-274, 17th European MPI Users' Group Meeting, EuroMPI 2010, Stuttgart, Germany, 09/12/10. https://doi.org/10.1007/978-3-642-15646-5_28

Locality and topology aware intra-node communication among multicore CPUs. / Ma, Teng; Bosilca, George; Bouteiller, Aurelien et al.
Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. 2010. p. 265-274 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6305 LNCS).

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

TY - GEN

T1 - Locality and topology aware intra-node communication among multicore CPUs

AU - Ma, Teng

AU - Bosilca, George

AU - Bouteiller, Aurelien

AU - Dongarra, Jack J.

PY - 2010

Y1 - 2010

N2 - A major trend in HPC is the escalation toward manycore, where systems are composed of shared memory nodes featuring numerous processing units. Unfortunately, with scale comes complexity, here in the form of non-uniform memory accesses and cache hierarchies. For most HPC applications, harnessing the power of multicores is hindered by the topology oblivious tuning of the MPI library. In this paper, we propose a framework to tune every type of shared memory communications according to locality and topology. An implementation inside Open MPI is evaluated experimentally and demonstrates significant speedups compared to vanilla Open MPI and MPICH2.

AB - A major trend in HPC is the escalation toward manycore, where systems are composed of shared memory nodes featuring numerous processing units. Unfortunately, with scale comes complexity, here in the form of non-uniform memory accesses and cache hierarchies. For most HPC applications, harnessing the power of multicores is hindered by the topology oblivious tuning of the MPI library. In this paper, we propose a framework to tune every type of shared memory communications according to locality and topology. An implementation inside Open MPI is evaluated experimentally and demonstrates significant speedups compared to vanilla Open MPI and MPICH2.

UR - http://www.scopus.com/inward/record.url?scp=78149268675&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15646-5_28

DO - 10.1007/978-3-642-15646-5_28

M3 - Conference contribution

AN - SCOPUS:78149268675

SN - 3642156452

SN - 9783642156458

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

SP - 265

EP - 274

BT - Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings

T2 - 17th European MPI Users' Group Meeting, EuroMPI 2010

Y2 - 12 September 2010 through 15 September 2010

ER -

Ma T, Bosilca G, Bouteiller A, Dongarra JJ. Locality and topology aware intra-node communication among multicore CPUs. In Recent Advances in the Message Passing Interface - 17th European MPI Users' Group Meeting, EuroMPI 2010, Proceedings. 2010. p. 265-274. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-15646-5_28

Locality and topology aware intra-node communication among multicore CPUs

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this