• DocumentCode
    2065967
  • Title

    Incremental 3D NoC synthesis based on physical-aware router merging algorithm

  • Author

    Yuanyuan Li ; Ning Xu ; Yuchun Ma ; Jinian Bian

  • Author_Institution
    Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China
  • fYear
    2013
  • fDate
    28-31 Oct. 2013
  • Firstpage
    1
  • Lastpage
    4
  • Abstract
    3D NoC (Network on Chip) technology has opened a new opportunity for the development of Network on Chip, which is a promising solution by distributing the network on multiple silicon layers. Comparing to the 2D designs, 3D NoC synthesis becomes even more complicated with stacking layers. The power consumption and network resources should be optimized with the consideration of the third dimension especially with Through-Silicon Vias (TSVs) among different layers. In this paper, by focusing on the network topology generation, we propose an incremental algorithm to optimize the network by merging the adjacent routers properly in 3D designs, so that many physical design issues could be optimized at the same time. Experimental results show that our algorithm can on average achieve 33% reduction in power consumption and 41% reduction in network resource area over the original 3D implementation, and the number of TSVs would be reduced by 11% on average.
  • Keywords
    integrated circuit design; network routing; network topology; network-on-chip; three-dimensional integrated circuits; incremental 3D NoC synthesis; incremental algorithm; network on chip technology; network topology generation; physical aware router merging algorithm; through silicon vias; Algorithm design and analysis; Bandwidth; Layout; Merging; Power demand; Three-dimensional displays; Topology;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    ASIC (ASICON), 2013 IEEE 10th International Conference on
  • Conference_Location
    Shenzhen
  • ISSN
    2162-7541
  • Print_ISBN
    978-1-4673-6415-7
  • Type

    conf

  • DOI
    10.1109/ASICON.2013.6811920
  • Filename
    6811920