• DocumentCode
    848494
  • Title

    Exploring architectural solutions for energy optimisations in bus-based system-on-chip

  • Author

    Srinivasan, S. ; Li, L. ; Ruggiero, M. ; Angiolini, F. ; Vijaykrishnan, N. ; Benini, L.

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Pennsylvania State Univ., University Park, PA
  • Volume
    2
  • Issue
    5
  • fYear
    2008
  • fDate
    9/1/2008 12:00:00 AM
  • Firstpage
    347
  • Lastpage
    354
  • Abstract
    System-on-chip (SoC) architectures have emerged as ubiquitous option for computation-intensive applications because of the tremendous flexibility provided by them with respect to design reuse and extendibility. Single shared bus architectures have been popularly used as communication channel in such on-chip architectures, by connecting various modules in such systems. However, increasing levels of integration and the number of components connected to the bus render such architectures infeasible. Consequently, such single shared bus architectures fail to scale well with both performance and power aspects. The authors demonstrate here a way to perform memory and bus partitioning and allocate variable frequencies to different bus segments to reduce the power consumption of the system without affecting the performance. The authors use an evolutionary algorithm followed by an iterative search-based frequency allocation algorithm to solve the problem. The effectiveness of the proposed results is validated on a SystemC-based cycle-accurate bus-based SoC simulator.
  • Keywords
    evolutionary computation; hardware description languages; iterative methods; logic design; system-on-chip; SystemC-based cycle-accurate bus; bus-based system-on-chip; computation-intensive applications; energy optimisations; evolutionary algorithm; iterative search-based frequency allocation;
  • fLanguage
    English
  • Journal_Title
    Computers & Digital Techniques, IET
  • Publisher
    iet
  • ISSN
    1751-8601
  • Type

    jour

  • DOI
    10.1049/iet-cdt:20070063
  • Filename
    4609371