Fast broadcast by the divide-and-conquer algorithm

Dongyoung Kim; Dongseung Kim

doi:10.1109/CLUSTR.2004.1392653

Fast broadcast by the divide-and-conquer algorithm

Dongyoung Kim, Dongseung Kim

GREEN SCHOOL (Graduate School of Energy and Environment)

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

3 Citations (Scopus)

Abstract

The features of a novel O(m) broadcast algorithm, independent of the number of processors involved, devised by using divided-and-conquer algorithm are discussed. Collective communication functions including the broadcast in cluster computers usually take O(m log p) time in propagating the size-m message to P processors. The overall time T(M,K) for the message broadcast of size M on K processors can be computed directly from the algorithm with T(M/2,K/2) and 2t(M/2), where t(M/2) is the point-to-point communication time of the message of size M/2 between arbitrary two processors. It is pointed that the new broadcast algorithm will be much faster if there are many participating processors for the broadcast and the message is large.

Original language	English
Title of host publication	2004 IEEE International Conference on Cluster Computing, ICCC 2004
Pages	487-488
Number of pages	2
DOIs	https://doi.org/10.1109/CLUSTR.2004.1392653
Publication status	Published - 2004
Event	2004 IEEE International Conference on Cluster Computing, ICCC 2004 - San Diego, CA, United States Duration: 2004 Sept 20 → 2004 Sept 23

Publication series

Name	Proceedings - IEEE International Conference on Cluster Computing, ICCC
ISSN (Print)	1552-5244

Other

Other	2004 IEEE International Conference on Cluster Computing, ICCC 2004
Country/Territory	United States
City	San Diego, CA
Period	04/9/20 → 04/9/23

ASJC Scopus subject areas

Engineering(all)

Access to Document

10.1109/CLUSTR.2004.1392653

Cite this

Kim, D & Kim, D 2004, Fast broadcast by the divide-and-conquer algorithm. in 2004 IEEE International Conference on Cluster Computing, ICCC 2004. Proceedings - IEEE International Conference on Cluster Computing, ICCC, pp. 487-488, 2004 IEEE International Conference on Cluster Computing, ICCC 2004, San Diego, CA, United States, 04/9/20. https://doi.org/10.1109/CLUSTR.2004.1392653

@inproceedings{3691df60010244f2938ff9e3528a1956,

title = "Fast broadcast by the divide-and-conquer algorithm",

abstract = "The features of a novel O(m) broadcast algorithm, independent of the number of processors involved, devised by using divided-and-conquer algorithm are discussed. Collective communication functions including the broadcast in cluster computers usually take O(m log p) time in propagating the size-m message to P processors. The overall time T(M,K) for the message broadcast of size M on K processors can be computed directly from the algorithm with T(M/2,K/2) and 2t(M/2), where t(M/2) is the point-to-point communication time of the message of size M/2 between arbitrary two processors. It is pointed that the new broadcast algorithm will be much faster if there are many participating processors for the broadcast and the message is large.",

author = "Dongyoung Kim and Dongseung Kim",

year = "2004",

doi = "10.1109/CLUSTR.2004.1392653",

language = "English",

isbn = "0780386949",

series = "Proceedings - IEEE International Conference on Cluster Computing, ICCC",

pages = "487--488",

booktitle = "2004 IEEE International Conference on Cluster Computing, ICCC 2004",

note = "2004 IEEE International Conference on Cluster Computing, ICCC 2004 ; Conference date: 20-09-2004 Through 23-09-2004",

}

TY - GEN

T1 - Fast broadcast by the divide-and-conquer algorithm

AU - Kim, Dongyoung

AU - Kim, Dongseung

PY - 2004

Y1 - 2004

N2 - The features of a novel O(m) broadcast algorithm, independent of the number of processors involved, devised by using divided-and-conquer algorithm are discussed. Collective communication functions including the broadcast in cluster computers usually take O(m log p) time in propagating the size-m message to P processors. The overall time T(M,K) for the message broadcast of size M on K processors can be computed directly from the algorithm with T(M/2,K/2) and 2t(M/2), where t(M/2) is the point-to-point communication time of the message of size M/2 between arbitrary two processors. It is pointed that the new broadcast algorithm will be much faster if there are many participating processors for the broadcast and the message is large.

AB - The features of a novel O(m) broadcast algorithm, independent of the number of processors involved, devised by using divided-and-conquer algorithm are discussed. Collective communication functions including the broadcast in cluster computers usually take O(m log p) time in propagating the size-m message to P processors. The overall time T(M,K) for the message broadcast of size M on K processors can be computed directly from the algorithm with T(M/2,K/2) and 2t(M/2), where t(M/2) is the point-to-point communication time of the message of size M/2 between arbitrary two processors. It is pointed that the new broadcast algorithm will be much faster if there are many participating processors for the broadcast and the message is large.

UR - http://www.scopus.com/inward/record.url?scp=20444505218&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=20444505218&partnerID=8YFLogxK

U2 - 10.1109/CLUSTR.2004.1392653

DO - 10.1109/CLUSTR.2004.1392653

M3 - Conference contribution

AN - SCOPUS:20444505218

SN - 0780386949

T3 - Proceedings - IEEE International Conference on Cluster Computing, ICCC

SP - 487

EP - 488

BT - 2004 IEEE International Conference on Cluster Computing, ICCC 2004

T2 - 2004 IEEE International Conference on Cluster Computing, ICCC 2004

Y2 - 20 September 2004 through 23 September 2004

ER -

Fast broadcast by the divide-and-conquer algorithm

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this