• DocumentCode
    3204965
  • Title

    Dominant representations: a paradigm for mapping parallel computations

  • Author

    Mohindra, Ajay ; Yalamanchili, Sudhakar

  • Author_Institution
    Georgia Inst. of Technol., Atlanta, GA, USA
  • fYear
    1992
  • fDate
    23-26 Mar 1992
  • Firstpage
    67
  • Lastpage
    71
  • Abstract
    Proposes a simple paradigm for constructing heuristics for the static assignment of parallel programs onto asynchronous, distributed memory, multiprocessor architectures. The proposed paradigm involves capturing the dominant computation and communication components of an application and using this relatively simpler program representation to determine an assignment. Thus, the mapping problem is reduced from its most general form to a simpler form which often has optimal solutions
  • Keywords
    distributed memory systems; parallel architectures; parallel programming; distributed memory; dominant representations; mapping problem; multiprocessor architectures; parallel programs; static assignment; Bandwidth; Computational efficiency; Computer architecture; Concurrent computing; Distributed computing; Educational institutions; Laboratories; Memory architecture; Runtime; Tree graphs;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Parallel Processing Symposium, 1992. Proceedings., Sixth International
  • Conference_Location
    Beverly Hills, CA
  • Print_ISBN
    0-8186-2672-0
  • Type

    conf

  • DOI
    10.1109/IPPS.1992.223070
  • Filename
    223070