• DocumentCode
    244019
  • Title

    A Graph-Based 3D IC Partitioning Technique

  • Author

    Banerjee, Sean ; Majumder, Subhashis ; Bhattacharya, Bhargab B.

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Heritage Inst. of Technol., Kolkata, India
  • fYear
    2014
  • fDate
    9-11 July 2014
  • Firstpage
    613
  • Lastpage
    618
  • Abstract
    Netlist partitioning is an important part of the physical design of 3D IC chips. Each subcircuit corresponding to a partition is subsequently assigned to a suitable device layer in the design phase. This paper proposes a netlist partitioning technique that intends to minimize the number of inter-layer interconnections while maintaining the area constraints. This, in turn, will minimize the area and cost associated with the Through-Silicon Vias (TSVs) needed in the design. The proposed method starts with an BFS-based initial solution and then improves iteratively using a heuristic. Experimental results demonstrate that by reassigning some modules to other layers, our algorithm could achieve up to 45% reduction in the number of TSVs on several benchmark circuits compared to earlier approaches. The resulting increase in floor area due to movement of modules a cross layers, is almost compensated by the decrease in TSV-area. Thus while satisfying the area-constraints, it allows us to reduce the number of TSVs as well as the IR-drop and delay associated with the vias.
  • Keywords
    integrated circuit design; integrated circuit interconnections; iterative methods; three-dimensional integrated circuits; area constraints; benchmark circuits; graph-based 3D IC partitioning technique; interlayer interconnections; netlist partitioning technique; through silicon vias; Algorithm design and analysis; Benchmark testing; Partitioning algorithms; Three-dimensional displays; Through-silicon vias; Time complexity; 3D IC; Through Silicon Vias (TSV); netlist partitioning;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    VLSI (ISVLSI), 2014 IEEE Computer Society Annual Symposium on
  • Conference_Location
    Tampa, FL
  • Print_ISBN
    978-1-4799-3763-9
  • Type

    conf

  • DOI
    10.1109/ISVLSI.2014.82
  • Filename
    6903432