Value-based distributions in fortran D

Reinhard V. Hanxleden; Ken Kennedy; Joel Saltz

doi:10.1007/3-540-57981-8_156

Value-based distributions in fortran D

Reinhard V. Hanxleden
, Ken Kennedy
, Joel Saltz

Biomedical Informatics

Stony Brook University

Rice University

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

1 Scopus citations

Abstract

Compiling irregular applications written in a data-parallel language, like High Performance Fortran, presents a challenging problem of growing importance. One difficulty with irregular problems is the lack of access locality when distributing data and computation naively, based on simple functions of array and iteration indices. Value-based distributions extend index-based distributions to conveniently express for example the locality characteristics of an underlying physical problem-. This gives the compiler an opportunity to improve both inter- and intraprocessor locality, resulting in better performance and scalability, even when this locality is not reflected by the sequential data structures and access patterns.

Original language	English
Title of host publication	High-Performance Computing and Networking
Subtitle of host publication	Networking and Tools - International Conference and Exhibition, Proceedings
Editors	Wolfgang Gentzsch, Uwe Harms
Publisher	Springer Verlag
Pages	434-440
Number of pages	7
ISBN (Print)	9783540579816
DOIs	https://doi.org/10.1007/3-540-57981-8_156
State	Published - 1994
Event	International Conference and Exhibition on High-Performance Computing and Networking, HPCN 1994 - Munich, Germany Duration: Apr 18 1994 → Apr 20 1994

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	797 LNCS

Conference

Conference	International Conference and Exhibition on High-Performance Computing and Networking, HPCN 1994
Country/Territory	Germany
City	Munich
Period	04/18/94 → 04/20/94

Access to Document

10.1007/3-540-57981-8_156

Cite this

Hanxleden, R. V., Kennedy, K., & Saltz, J. (1994). Value-based distributions in fortran D. In W. Gentzsch, & U. Harms (Eds.), High-Performance Computing and Networking: Networking and Tools - International Conference and Exhibition, Proceedings (pp. 434-440). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 797 LNCS). Springer Verlag. https://doi.org/10.1007/3-540-57981-8_156

Hanxleden, Reinhard V. ; Kennedy, Ken ; Saltz, Joel. / Value-based distributions in fortran D. High-Performance Computing and Networking: Networking and Tools - International Conference and Exhibition, Proceedings. editor / Wolfgang Gentzsch ; Uwe Harms. Springer Verlag, 1994. pp. 434-440 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{40ee1d386f3d4cc58c448be8c7e20845,

title = "Value-based distributions in fortran D",

abstract = "Compiling irregular applications written in a data-parallel language, like High Performance Fortran, presents a challenging problem of growing importance. One difficulty with irregular problems is the lack of access locality when distributing data and computation naively, based on simple functions of array and iteration indices. Value-based distributions extend index-based distributions to conveniently express for example the locality characteristics of an underlying physical problem-. This gives the compiler an opportunity to improve both inter- and intraprocessor locality, resulting in better performance and scalability, even when this locality is not reflected by the sequential data structures and access patterns.",

author = "Hanxleden, \{Reinhard V.\} and Ken Kennedy and Joel Saltz",

note = "Publisher Copyright: {\textcopyright} 1994, Springer Verlag. All rights reserved.; International Conference and Exhibition on High-Performance Computing and Networking, HPCN 1994 ; Conference date: 18-04-1994 Through 20-04-1994",

year = "1994",

doi = "10.1007/3-540-57981-8\_156",

language = "English",

isbn = "9783540579816",

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

publisher = "Springer Verlag",

pages = "434--440",

editor = "Wolfgang Gentzsch and Uwe Harms",

booktitle = "High-Performance Computing and Networking",

}

Hanxleden, RV, Kennedy, K & Saltz, J 1994, Value-based distributions in fortran D. in W Gentzsch & U Harms (eds), High-Performance Computing and Networking: Networking and Tools - International Conference and Exhibition, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 797 LNCS, Springer Verlag, pp. 434-440, International Conference and Exhibition on High-Performance Computing and Networking, HPCN 1994, Munich, Germany, 04/18/94. https://doi.org/10.1007/3-540-57981-8_156

Value-based distributions in fortran D. / Hanxleden, Reinhard V.; Kennedy, Ken; Saltz, Joel.
High-Performance Computing and Networking: Networking and Tools - International Conference and Exhibition, Proceedings. ed. / Wolfgang Gentzsch; Uwe Harms. Springer Verlag, 1994. p. 434-440 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 797 LNCS).

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

TY - GEN

T1 - Value-based distributions in fortran D

AU - Hanxleden, Reinhard V.

AU - Kennedy, Ken

AU - Saltz, Joel

PY - 1994

Y1 - 1994

N2 - Compiling irregular applications written in a data-parallel language, like High Performance Fortran, presents a challenging problem of growing importance. One difficulty with irregular problems is the lack of access locality when distributing data and computation naively, based on simple functions of array and iteration indices. Value-based distributions extend index-based distributions to conveniently express for example the locality characteristics of an underlying physical problem-. This gives the compiler an opportunity to improve both inter- and intraprocessor locality, resulting in better performance and scalability, even when this locality is not reflected by the sequential data structures and access patterns.

AB - Compiling irregular applications written in a data-parallel language, like High Performance Fortran, presents a challenging problem of growing importance. One difficulty with irregular problems is the lack of access locality when distributing data and computation naively, based on simple functions of array and iteration indices. Value-based distributions extend index-based distributions to conveniently express for example the locality characteristics of an underlying physical problem-. This gives the compiler an opportunity to improve both inter- and intraprocessor locality, resulting in better performance and scalability, even when this locality is not reflected by the sequential data structures and access patterns.

UR - https://www.scopus.com/pages/publications/21344498237

U2 - 10.1007/3-540-57981-8_156

DO - 10.1007/3-540-57981-8_156

M3 - Conference contribution

SN - 9783540579816

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

SP - 434

EP - 440

BT - High-Performance Computing and Networking

A2 - Gentzsch, Wolfgang

A2 - Harms, Uwe

PB - Springer Verlag

T2 - International Conference and Exhibition on High-Performance Computing and Networking, HPCN 1994

Y2 - 18 April 1994 through 20 April 1994

ER -

Hanxleden RV, Kennedy K, Saltz J. Value-based distributions in fortran D. In Gentzsch W, Harms U, editors, High-Performance Computing and Networking: Networking and Tools - International Conference and Exhibition, Proceedings. Springer Verlag. 1994. p. 434-440. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/3-540-57981-8_156

Value-based distributions in fortran D

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this