Profile guided MPI protocol selection for point-to-point communication calls

Aniruddha Marathe, David K Lowenthal, Zheng Gu, Matthew Small, Xin Yuan

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

Improving communication performance is critical to achieving high performance in message-passing programs. Designing new, efficient protocols to realize point-to-point and collective communication operations has therefore been an active area of research. However, the best protocol for a given communication routine is both application and architecture specific. This paper contributes a new method of selection of the optimal protocol for a given point-to-point communication pair. Our technique analyzes the MPI communication call profile of an application and uses a computation and communication model we have developed to choose the proper protocol for each communication phase. We have applied our system to MPI applications such as CG, Sweep3D and Sparse Matrix multiplication, as well as synthetic applications. Our scheme yields an improvement in total execution time of up to 20% compared to MVAPICH2 and up to 3.2% compared to the best, highly optimized communication protocol for the real applications. Furthermore, experiments on the synthetic applications show that the savings can be much more pronounced.

Original languageEnglish (US)
Title of host publicationIEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum
Pages733-739
Number of pages7
DOIs
StatePublished - 2011
Event25th IEEE International Parallel and Distributed Processing Symposium, Workshops and Phd Forum, IPDPSW 2011 - Anchorage, AK, United States
Duration: May 16 2011May 20 2011

Other

Other25th IEEE International Parallel and Distributed Processing Symposium, Workshops and Phd Forum, IPDPSW 2011
CountryUnited States
CityAnchorage, AK
Period5/16/115/20/11

Fingerprint

Network protocols
Communication
Collective Communication
Matrix multiplication
Communication Protocol
Sparse matrix
Message Passing
Message passing
Execution Time
Profile
High Performance
Choose
Experiment
Experiments
Model

Keywords

  • MPI
  • Point-to-point communication
  • Protocol selection

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Software
  • Theoretical Computer Science

Cite this

Marathe, A., Lowenthal, D. K., Gu, Z., Small, M., & Yuan, X. (2011). Profile guided MPI protocol selection for point-to-point communication calls. In IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum (pp. 733-739). [6008844] https://doi.org/10.1109/IPDPS.2011.215

Profile guided MPI protocol selection for point-to-point communication calls. / Marathe, Aniruddha; Lowenthal, David K; Gu, Zheng; Small, Matthew; Yuan, Xin.

IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum. 2011. p. 733-739 6008844.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Marathe, A, Lowenthal, DK, Gu, Z, Small, M & Yuan, X 2011, Profile guided MPI protocol selection for point-to-point communication calls. in IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum., 6008844, pp. 733-739, 25th IEEE International Parallel and Distributed Processing Symposium, Workshops and Phd Forum, IPDPSW 2011, Anchorage, AK, United States, 5/16/11. https://doi.org/10.1109/IPDPS.2011.215
Marathe A, Lowenthal DK, Gu Z, Small M, Yuan X. Profile guided MPI protocol selection for point-to-point communication calls. In IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum. 2011. p. 733-739. 6008844 https://doi.org/10.1109/IPDPS.2011.215
Marathe, Aniruddha ; Lowenthal, David K ; Gu, Zheng ; Small, Matthew ; Yuan, Xin. / Profile guided MPI protocol selection for point-to-point communication calls. IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum. 2011. pp. 733-739
@inproceedings{56d81af8d60c400c91d5651025ec778b,
title = "Profile guided MPI protocol selection for point-to-point communication calls",
abstract = "Improving communication performance is critical to achieving high performance in message-passing programs. Designing new, efficient protocols to realize point-to-point and collective communication operations has therefore been an active area of research. However, the best protocol for a given communication routine is both application and architecture specific. This paper contributes a new method of selection of the optimal protocol for a given point-to-point communication pair. Our technique analyzes the MPI communication call profile of an application and uses a computation and communication model we have developed to choose the proper protocol for each communication phase. We have applied our system to MPI applications such as CG, Sweep3D and Sparse Matrix multiplication, as well as synthetic applications. Our scheme yields an improvement in total execution time of up to 20{\%} compared to MVAPICH2 and up to 3.2{\%} compared to the best, highly optimized communication protocol for the real applications. Furthermore, experiments on the synthetic applications show that the savings can be much more pronounced.",
keywords = "MPI, Point-to-point communication, Protocol selection",
author = "Aniruddha Marathe and Lowenthal, {David K} and Zheng Gu and Matthew Small and Xin Yuan",
year = "2011",
doi = "10.1109/IPDPS.2011.215",
language = "English (US)",
isbn = "9780769543857",
pages = "733--739",
booktitle = "IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum",

}

TY - GEN

T1 - Profile guided MPI protocol selection for point-to-point communication calls

AU - Marathe, Aniruddha

AU - Lowenthal, David K

AU - Gu, Zheng

AU - Small, Matthew

AU - Yuan, Xin

PY - 2011

Y1 - 2011

N2 - Improving communication performance is critical to achieving high performance in message-passing programs. Designing new, efficient protocols to realize point-to-point and collective communication operations has therefore been an active area of research. However, the best protocol for a given communication routine is both application and architecture specific. This paper contributes a new method of selection of the optimal protocol for a given point-to-point communication pair. Our technique analyzes the MPI communication call profile of an application and uses a computation and communication model we have developed to choose the proper protocol for each communication phase. We have applied our system to MPI applications such as CG, Sweep3D and Sparse Matrix multiplication, as well as synthetic applications. Our scheme yields an improvement in total execution time of up to 20% compared to MVAPICH2 and up to 3.2% compared to the best, highly optimized communication protocol for the real applications. Furthermore, experiments on the synthetic applications show that the savings can be much more pronounced.

AB - Improving communication performance is critical to achieving high performance in message-passing programs. Designing new, efficient protocols to realize point-to-point and collective communication operations has therefore been an active area of research. However, the best protocol for a given communication routine is both application and architecture specific. This paper contributes a new method of selection of the optimal protocol for a given point-to-point communication pair. Our technique analyzes the MPI communication call profile of an application and uses a computation and communication model we have developed to choose the proper protocol for each communication phase. We have applied our system to MPI applications such as CG, Sweep3D and Sparse Matrix multiplication, as well as synthetic applications. Our scheme yields an improvement in total execution time of up to 20% compared to MVAPICH2 and up to 3.2% compared to the best, highly optimized communication protocol for the real applications. Furthermore, experiments on the synthetic applications show that the savings can be much more pronounced.

KW - MPI

KW - Point-to-point communication

KW - Protocol selection

UR - http://www.scopus.com/inward/record.url?scp=83455166564&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=83455166564&partnerID=8YFLogxK

U2 - 10.1109/IPDPS.2011.215

DO - 10.1109/IPDPS.2011.215

M3 - Conference contribution

AN - SCOPUS:83455166564

SN - 9780769543857

SP - 733

EP - 739

BT - IEEE International Symposium on Parallel and Distributed Processing Workshops and Phd Forum

ER -