• DocumentCode
    2604267
  • Title

    Performance analysis and optimization for homogenous multi-core system based on 3D Torus Network on Chip

  • Author

    Jiao, Jiajia ; Fu, Yuzhuo ; Liu, Ting ; Wang, Han ; Han, Xing ; Wang, Jiafang

  • Author_Institution
    Sch. of Microelectron., Shanghai Jiao Tong Univ., Shanghai, China
  • fYear
    2010
  • fDate
    20-23 June 2010
  • Firstpage
    313
  • Lastpage
    316
  • Abstract
    3D IC technology drives Network-On-Chip (NoC) design on towards 3D trend and relevant multi-core system further development. However, most recent researches still focus on the fundamental 3D Mesh structure and have no convincible traffic pattern models in realistic applications. In this paper, we p resent a complete design framework of a Distributed Shared Memory homogenous multi-core system based on 3D Torus interconnects. A Gray code routing mechanism is also proposed to solve deadlock. In addition, we evaluate the 3D Torus system with both Synthetic traffic (uniform and transpose pattern) and real parallel multi-thread benchmark. Comparing with 3D Mesh, our experimental results demonstrate a significant reduction in the average packet latency and higher throughout put as well as a major improvement of execution cycles in typical thermal-aware IP mapping methods in 3D Torus multi-core system.
  • Keywords
    distributed shared memory systems; integrated circuit design; multi-threading; multiprocessing systems; network routing; network-on-chip; performance evaluation; 3D IC technology; 3D mesh structure; 3D torus interconnects; 3D torus multicore system; 3D torus network on chip; 3D torus system; NoC design; distributed shared memory homogenous multicore system; gray code routing mechanism; network-on-chip design; optimization; packet latency; performance analysis; real parallel multithread benchmark; synthetic traffic; thermal-aware IP mapping methods; traffic pattern models; IP networks; Reflective binary codes; Routing; System recovery; Three dimensional displays; Topology; Traffic control; 3D Torus; Deadlock-free routing; Distributed shared memory; Homogenous multi-core; Thread-level Parallelism;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    NEWCAS Conference (NEWCAS), 2010 8th IEEE International
  • Conference_Location
    Montreal, QC
  • Print_ISBN
    978-1-4244-6806-5
  • Electronic_ISBN
    978-1-4244-6804-1
  • Type

    conf

  • DOI
    10.1109/NEWCAS.2010.5604033
  • Filename
    5604033