Parallel matrix transpose algorithms on distributed memory concurrent computers

Jaeyoung Choi; J. J. Dongarra; D. W. Walker

doi:10.1109/SPLC.1993.365559

Parallel matrix transpose algorithms on distributed memory concurrent computers

Jaeyoung Choi, J. J. Dongarra, D. W. Walker

Computer Science and Mathematics Div

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

11 Scopus citations

Abstract

This paper describes parallel matrix transpose algorithms on distributed memory concurrent processors. We assume that the matrix is distributed over a P×Q processor template with a block scattered data distribution. P, Q, and the block size can be arbitrary, so the algorithms have wide applicability. The algorithms make use of non-blocking, point-to-point communication between processors. The use of nonblocking communication allows a processor to overlap the messages that it sends to different processors, thereby avoiding unnecessary synchronization. Combined with the matrix multiplication routine, C=A·B, the algorithms are used to compute parallel multiplications of transposed matrices, C=A^T·B^T, in the PUMMA package. Details of the parallel implementation of the algorithms are given, and results are presented for runs on the Intel Touchstone Delta computer.

Original language	English
Title of host publication	Proceedings of Scalable Parallel Libraries Conference, SPLC 1993
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	245-252
Number of pages	8
ISBN (Electronic)	0818649801, 9780818649806
DOIs	https://doi.org/10.1109/SPLC.1993.365559
State	Published - 1993
Event	1993 Scalable Parallel Libraries Conference, SPLC 1993 - Mississippi State, United States Duration: Oct 6 1993 → Oct 8 1993

Publication series

Name	Proceedings of Scalable Parallel Libraries Conference, SPLC 1993

Conference

Conference	1993 Scalable Parallel Libraries Conference, SPLC 1993
Country/Territory	United States
City	Mississippi State
Period	10/6/93 → 10/8/93

Access to Document

10.1109/SPLC.1993.365559

Cite this

Choi, J., Dongarra, J. J., & Walker, D. W. (1993). Parallel matrix transpose algorithms on distributed memory concurrent computers. In Proceedings of Scalable Parallel Libraries Conference, SPLC 1993 (pp. 245-252). Article 365559 (Proceedings of Scalable Parallel Libraries Conference, SPLC 1993). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SPLC.1993.365559

@inproceedings{729f53ae6cc7460eb18a1fee9b33da36,

title = "Parallel matrix transpose algorithms on distributed memory concurrent computers",

abstract = "This paper describes parallel matrix transpose algorithms on distributed memory concurrent processors. We assume that the matrix is distributed over a P×Q processor template with a block scattered data distribution. P, Q, and the block size can be arbitrary, so the algorithms have wide applicability. The algorithms make use of non-blocking, point-to-point communication between processors. The use of nonblocking communication allows a processor to overlap the messages that it sends to different processors, thereby avoiding unnecessary synchronization. Combined with the matrix multiplication routine, C=A·B, the algorithms are used to compute parallel multiplications of transposed matrices, C=AT·BT, in the PUMMA package. Details of the parallel implementation of the algorithms are given, and results are presented for runs on the Intel Touchstone Delta computer.",

author = "Jaeyoung Choi and Dongarra, {J. J.} and Walker, {D. W.}",

note = "Publisher Copyright: {\textcopyright} 1994 IEEE.; 1993 Scalable Parallel Libraries Conference, SPLC 1993 ; Conference date: 06-10-1993 Through 08-10-1993",

year = "1993",

doi = "10.1109/SPLC.1993.365559",

language = "English",

series = "Proceedings of Scalable Parallel Libraries Conference, SPLC 1993",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "245--252",

booktitle = "Proceedings of Scalable Parallel Libraries Conference, SPLC 1993",

}

Choi, J, Dongarra, JJ & Walker, DW 1993, Parallel matrix transpose algorithms on distributed memory concurrent computers. in Proceedings of Scalable Parallel Libraries Conference, SPLC 1993., 365559, Proceedings of Scalable Parallel Libraries Conference, SPLC 1993, Institute of Electrical and Electronics Engineers Inc., pp. 245-252, 1993 Scalable Parallel Libraries Conference, SPLC 1993, Mississippi State, United States, 10/6/93. https://doi.org/10.1109/SPLC.1993.365559

Parallel matrix transpose algorithms on distributed memory concurrent computers. / Choi, Jaeyoung; Dongarra, J. J.; Walker, D. W.
Proceedings of Scalable Parallel Libraries Conference, SPLC 1993. Institute of Electrical and Electronics Engineers Inc., 1993. p. 245-252 365559 (Proceedings of Scalable Parallel Libraries Conference, SPLC 1993).

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

TY - GEN

T1 - Parallel matrix transpose algorithms on distributed memory concurrent computers

AU - Choi, Jaeyoung

AU - Dongarra, J. J.

AU - Walker, D. W.

PY - 1993

Y1 - 1993

N2 - This paper describes parallel matrix transpose algorithms on distributed memory concurrent processors. We assume that the matrix is distributed over a P×Q processor template with a block scattered data distribution. P, Q, and the block size can be arbitrary, so the algorithms have wide applicability. The algorithms make use of non-blocking, point-to-point communication between processors. The use of nonblocking communication allows a processor to overlap the messages that it sends to different processors, thereby avoiding unnecessary synchronization. Combined with the matrix multiplication routine, C=A·B, the algorithms are used to compute parallel multiplications of transposed matrices, C=AT·BT, in the PUMMA package. Details of the parallel implementation of the algorithms are given, and results are presented for runs on the Intel Touchstone Delta computer.

AB - This paper describes parallel matrix transpose algorithms on distributed memory concurrent processors. We assume that the matrix is distributed over a P×Q processor template with a block scattered data distribution. P, Q, and the block size can be arbitrary, so the algorithms have wide applicability. The algorithms make use of non-blocking, point-to-point communication between processors. The use of nonblocking communication allows a processor to overlap the messages that it sends to different processors, thereby avoiding unnecessary synchronization. Combined with the matrix multiplication routine, C=A·B, the algorithms are used to compute parallel multiplications of transposed matrices, C=AT·BT, in the PUMMA package. Details of the parallel implementation of the algorithms are given, and results are presented for runs on the Intel Touchstone Delta computer.

UR - http://www.scopus.com/inward/record.url?scp=84947031567&partnerID=8YFLogxK

U2 - 10.1109/SPLC.1993.365559

DO - 10.1109/SPLC.1993.365559

M3 - Conference contribution

AN - SCOPUS:84947031567

T3 - Proceedings of Scalable Parallel Libraries Conference, SPLC 1993

SP - 245

EP - 252

BT - Proceedings of Scalable Parallel Libraries Conference, SPLC 1993

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 1993 Scalable Parallel Libraries Conference, SPLC 1993

Y2 - 6 October 1993 through 8 October 1993

ER -

Parallel matrix transpose algorithms on distributed memory concurrent computers

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this