Data localization using loop aligned decomposition for macro-dataflow processing

Akimasa Yoshida, Hironori Kasahara

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

    Abstract

    This paper proposes a data-localization compilation scheme for Fortran macro-dataflow processing on a multiprocessor system with local memory and centralized shared memory. The data-localization scheme minimizes data transfer overhead for passing shared data among coarse-grain tasks composed of Doall loops and sequential loops by using local memory on each processor effectively. In this scheme, a compiler firstly partitions coarse-grain tasks like loops having data dependences among them and their data into multiple groups by a loop aligned decomposition so that data transfer among groups can be minimum. Secondly it generates dynamic scheduling routine which assigns decomposed tasks in a group to the same processor at run-time. Thirdly it generates parallel machine code to pass shared data inside the group through local memory. This compiler has been implemented for an multiprocessor system OSCAR having centralized shared memory and distributed shared memory in addition to local memory on each processor. Performance evaluation on OSCAR shows that macro-dataflow processing with the proposed data-localization scheme can reduce the execution time by 10% to 20% in average compared with macro-dataflow processing without data-localization.

    Original languageEnglish
    Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    PublisherSpringer Verlag
    Pages57-74
    Number of pages18
    Volume1239
    ISBN (Print)3540630910, 9783540630913
    DOIs
    Publication statusPublished - 1997
    Event9th International Workshop on Languages and Compilers for Parallel Computing, LCPC 1996 - San Jose, United States
    Duration: 1996 Aug 81996 Aug 10

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume1239
    ISSN (Print)03029743
    ISSN (Electronic)16113349

    Other

    Other9th International Workshop on Languages and Compilers for Parallel Computing, LCPC 1996
    CountryUnited States
    CitySan Jose
    Period96/8/896/8/10

    ASJC Scopus subject areas

    • Computer Science(all)
    • Theoretical Computer Science

    Fingerprint Dive into the research topics of 'Data localization using loop aligned decomposition for macro-dataflow processing'. Together they form a unique fingerprint.

  • Cite this

    Yoshida, A., & Kasahara, H. (1997). Data localization using loop aligned decomposition for macro-dataflow processing. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1239, pp. 57-74). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1239). Springer Verlag. https://doi.org/10.1007/BFb0017245