Parallel power-of-two FFTs on hypercubes

Mahn ling Woo, Rosemary Renaut

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

1 Citation (Scopus)

Abstract

Theoretical analysis of ordered power-of-two fast Fourier transforms (PO2FFTs) demonstrates that the most efficient algorithms for hypercube architectures may not always use nearest-neighbor communication. It is shown that an ordered PO2FFT can be obtained with just d interprocessor communications for any r if the restriction that all communications are distance one is removed. The two interprocessor communications requirements of this algorithm are determined. Packets of size N/2d+1 are transmitted in both of the ordered PO2FFT algorithms. The time complexity of both the distance one and the distance two algorithms is discussed.

Original languageEnglish (US)
Title of host publicationProc Supercomput 91
Place of PublicationPiscataway, NJ, United States
PublisherPubl by IEEE
Pages754-763
Number of pages10
ISBN (Print)0818621583
StatePublished - 1991
EventProceedings of Supercomputing '91 - Albuquerque, NM, USA
Duration: Nov 18 1991Nov 22 1991

Other

OtherProceedings of Supercomputing '91
CityAlbuquerque, NM, USA
Period11/18/9111/22/91

Fingerprint

Fast Fourier transforms
Communication

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Woo, M. L., & Renaut, R. (1991). Parallel power-of-two FFTs on hypercubes. In Proc Supercomput 91 (pp. 754-763). Piscataway, NJ, United States: Publ by IEEE.

Parallel power-of-two FFTs on hypercubes. / Woo, Mahn ling; Renaut, Rosemary.

Proc Supercomput 91. Piscataway, NJ, United States : Publ by IEEE, 1991. p. 754-763.

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

Woo, ML & Renaut, R 1991, Parallel power-of-two FFTs on hypercubes. in Proc Supercomput 91. Publ by IEEE, Piscataway, NJ, United States, pp. 754-763, Proceedings of Supercomputing '91, Albuquerque, NM, USA, 11/18/91.
Woo ML, Renaut R. Parallel power-of-two FFTs on hypercubes. In Proc Supercomput 91. Piscataway, NJ, United States: Publ by IEEE. 1991. p. 754-763
Woo, Mahn ling ; Renaut, Rosemary. / Parallel power-of-two FFTs on hypercubes. Proc Supercomput 91. Piscataway, NJ, United States : Publ by IEEE, 1991. pp. 754-763
@inproceedings{4a9f16ef1c3549608e103dc6d23976e3,
title = "Parallel power-of-two FFTs on hypercubes",
abstract = "Theoretical analysis of ordered power-of-two fast Fourier transforms (PO2FFTs) demonstrates that the most efficient algorithms for hypercube architectures may not always use nearest-neighbor communication. It is shown that an ordered PO2FFT can be obtained with just d interprocessor communications for any r if the restriction that all communications are distance one is removed. The two interprocessor communications requirements of this algorithm are determined. Packets of size N/2d+1 are transmitted in both of the ordered PO2FFT algorithms. The time complexity of both the distance one and the distance two algorithms is discussed.",
author = "Woo, {Mahn ling} and Rosemary Renaut",
year = "1991",
language = "English (US)",
isbn = "0818621583",
pages = "754--763",
booktitle = "Proc Supercomput 91",
publisher = "Publ by IEEE",

}

TY - GEN

T1 - Parallel power-of-two FFTs on hypercubes

AU - Woo, Mahn ling

AU - Renaut, Rosemary

PY - 1991

Y1 - 1991

N2 - Theoretical analysis of ordered power-of-two fast Fourier transforms (PO2FFTs) demonstrates that the most efficient algorithms for hypercube architectures may not always use nearest-neighbor communication. It is shown that an ordered PO2FFT can be obtained with just d interprocessor communications for any r if the restriction that all communications are distance one is removed. The two interprocessor communications requirements of this algorithm are determined. Packets of size N/2d+1 are transmitted in both of the ordered PO2FFT algorithms. The time complexity of both the distance one and the distance two algorithms is discussed.

AB - Theoretical analysis of ordered power-of-two fast Fourier transforms (PO2FFTs) demonstrates that the most efficient algorithms for hypercube architectures may not always use nearest-neighbor communication. It is shown that an ordered PO2FFT can be obtained with just d interprocessor communications for any r if the restriction that all communications are distance one is removed. The two interprocessor communications requirements of this algorithm are determined. Packets of size N/2d+1 are transmitted in both of the ordered PO2FFT algorithms. The time complexity of both the distance one and the distance two algorithms is discussed.

UR - http://www.scopus.com/inward/record.url?scp=0026275675&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026275675&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0026275675

SN - 0818621583

SP - 754

EP - 763

BT - Proc Supercomput 91

PB - Publ by IEEE

CY - Piscataway, NJ, United States

ER -