Serial combinators

“optimal” grains of parallelism

Paul Hudak, Benjamin Goldberg

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

Abstract

A method is described for translating a high-level functional program into combinators suitable for execution on multiprocessors with no shared memory. It is argued that the granularity of the standard set of fixed combinators is too fine, whereas the granularity of user-defined functions is too coarse. The notion of a serial combinator is introduced that in some sense has optimal granularity, and that takes into account pragmatic issues such as the complexity of expressions and communication costs between processors. The technique improves on the standard notion of super-combinators by making them larger to retain locality and improve efficiency, and by ensuring that they have no concurrent substructure that could result in lost parallelism. Simulation results demonstrate improved performance on both sequential and parallel computing models.

Original languageEnglish (US)
Title of host publicationFunctional Programming Languages and Computer Architecture
PublisherSpringer Verlag
Pages382-399
Number of pages18
Volume201 LNCS
ISBN (Print)9783540159759
DOIs
StatePublished - Jan 1 1985
Event2nd International Conference on Functional Programming Languages and Computer Architecture, 1985 - Nancy, France
Duration: Sep 16 1985Sep 19 1985

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume201 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other2nd International Conference on Functional Programming Languages and Computer Architecture, 1985
CountryFrance
CityNancy
Period9/16/859/19/85

Fingerprint

Granularity
Parallelism
Parallel processing systems
Communication Cost
Substructure
Parallel Computing
Shared Memory
Multiprocessor
Locality
Data storage equipment
Concurrent
Communication
Costs
Demonstrate
Simulation
Standards
Model

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Hudak, P., & Goldberg, B. (1985). Serial combinators: “optimal” grains of parallelism. In Functional Programming Languages and Computer Architecture (Vol. 201 LNCS, pp. 382-399). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 201 LNCS). Springer Verlag. https://doi.org/10.1007/3-540-15975-4_49

Serial combinators : “optimal” grains of parallelism. / Hudak, Paul; Goldberg, Benjamin.

Functional Programming Languages and Computer Architecture. Vol. 201 LNCS Springer Verlag, 1985. p. 382-399 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 201 LNCS).

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

Hudak, P & Goldberg, B 1985, Serial combinators: “optimal” grains of parallelism. in Functional Programming Languages and Computer Architecture. vol. 201 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 201 LNCS, Springer Verlag, pp. 382-399, 2nd International Conference on Functional Programming Languages and Computer Architecture, 1985, Nancy, France, 9/16/85. https://doi.org/10.1007/3-540-15975-4_49
Hudak P, Goldberg B. Serial combinators: “optimal” grains of parallelism. In Functional Programming Languages and Computer Architecture. Vol. 201 LNCS. Springer Verlag. 1985. p. 382-399. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/3-540-15975-4_49
Hudak, Paul ; Goldberg, Benjamin. / Serial combinators : “optimal” grains of parallelism. Functional Programming Languages and Computer Architecture. Vol. 201 LNCS Springer Verlag, 1985. pp. 382-399 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{066a63ac35af4fc083a754bea5c56d53,
title = "Serial combinators: “optimal” grains of parallelism",
abstract = "A method is described for translating a high-level functional program into combinators suitable for execution on multiprocessors with no shared memory. It is argued that the granularity of the standard set of fixed combinators is too fine, whereas the granularity of user-defined functions is too coarse. The notion of a serial combinator is introduced that in some sense has optimal granularity, and that takes into account pragmatic issues such as the complexity of expressions and communication costs between processors. The technique improves on the standard notion of super-combinators by making them larger to retain locality and improve efficiency, and by ensuring that they have no concurrent substructure that could result in lost parallelism. Simulation results demonstrate improved performance on both sequential and parallel computing models.",
author = "Paul Hudak and Benjamin Goldberg",
year = "1985",
month = "1",
day = "1",
doi = "10.1007/3-540-15975-4_49",
language = "English (US)",
isbn = "9783540159759",
volume = "201 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "382--399",
booktitle = "Functional Programming Languages and Computer Architecture",
address = "Germany",

}

TY - GEN

T1 - Serial combinators

T2 - “optimal” grains of parallelism

AU - Hudak, Paul

AU - Goldberg, Benjamin

PY - 1985/1/1

Y1 - 1985/1/1

N2 - A method is described for translating a high-level functional program into combinators suitable for execution on multiprocessors with no shared memory. It is argued that the granularity of the standard set of fixed combinators is too fine, whereas the granularity of user-defined functions is too coarse. The notion of a serial combinator is introduced that in some sense has optimal granularity, and that takes into account pragmatic issues such as the complexity of expressions and communication costs between processors. The technique improves on the standard notion of super-combinators by making them larger to retain locality and improve efficiency, and by ensuring that they have no concurrent substructure that could result in lost parallelism. Simulation results demonstrate improved performance on both sequential and parallel computing models.

AB - A method is described for translating a high-level functional program into combinators suitable for execution on multiprocessors with no shared memory. It is argued that the granularity of the standard set of fixed combinators is too fine, whereas the granularity of user-defined functions is too coarse. The notion of a serial combinator is introduced that in some sense has optimal granularity, and that takes into account pragmatic issues such as the complexity of expressions and communication costs between processors. The technique improves on the standard notion of super-combinators by making them larger to retain locality and improve efficiency, and by ensuring that they have no concurrent substructure that could result in lost parallelism. Simulation results demonstrate improved performance on both sequential and parallel computing models.

UR - http://www.scopus.com/inward/record.url?scp=85034988175&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85034988175&partnerID=8YFLogxK

U2 - 10.1007/3-540-15975-4_49

DO - 10.1007/3-540-15975-4_49

M3 - Conference contribution

SN - 9783540159759

VL - 201 LNCS

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

SP - 382

EP - 399

BT - Functional Programming Languages and Computer Architecture

PB - Springer Verlag

ER -