• DocumentCode
    1998057
  • Title

    Leveraging GPUs in Ab Initio Nuclear Physics Calculations

  • Author

    Oryspayev, Dossay ; Potter, Henrique ; Maris, Pieter ; Sosonkina, Masha ; Vary, James P. ; Binder, S. ; Calci, Angelo ; Langhammer, Joachim ; Roth, R.

  • Author_Institution
    U.S. DOE Ames Lab., Iowa State Univ., Ames, IA, USA
  • fYear
    2013
  • fDate
    20-24 May 2013
  • Firstpage
    1365
  • Lastpage
    1372
  • Abstract
    This paper describes initial steps to leverage accelerators, such as GPUs, in ab initio nuclear physics calculations. Specifically, parallel nuclear structure calculations performed by the MFDn package are considered with selected stages adapted for GPUs. This paper outlines the necessary steps to make MFDn utilize GPUs in its matrix construction stage. The experiments are presented to compare the multithreaded CPU-based version of this phase with the newly designed GPU-based version. Four-to ten-fold speedups were observed for the code executing on GPUs.
  • Keywords
    configuration interactions; graphics processing units; matrix algebra; multi-threading; multiprocessing systems; nuclear physics; GPU-based version; MFDn package; ab initio nuclear physics calculations; code execution; leverage accelerators; matrix construction stage; multithreaded CPU-based version; parallel nuclear structure calculations; Graphics processing units; Indexes; Message systems; Sparse matrices; Symmetric matrices; Vectors; Ab initio nuclear structure calculation; CUDA; GPU; Multithreaded execution;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Parallel and Distributed Processing Symposium Workshops & PhD Forum (IPDPSW), 2013 IEEE 27th International
  • Conference_Location
    Cambridge, MA
  • Print_ISBN
    978-0-7695-4979-8
  • Type

    conf

  • DOI
    10.1109/IPDPSW.2013.253
  • Filename
    6651029