Adaptable wire-length distribution with tunable occupation probability

  • Authors:
  • Shuhei Amakawa;Takumi Uezono;Takashi Sato;Kenichi Okada;Kazuya Masu

  • Affiliations:
  • Tokyo Institute of Technology, Yokohama, Japan;Tokyo Institute of Technology, Yokohama, Japan;Tokyo Institute of Technology, Yokohama, Japan;Tokyo Institute of Technology, Yokohama, Japan;Tokyo Institute of Technology, Yokohama, Japan

  • Venue:
  • Proceedings of the 2007 international workshop on System level interconnect prediction
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Advance in wire-length distribution models in recent years has made possible fairly accurate early-stage prediction of interconnect requirements. Many models, however, tend to have no tunable parameters and the user cannot change the shape of the distribution. If the shape of an actual distribution is different from what is expressible by the model in use, there is not much one can do. This paper proposes to let the user change the shape of distribution. It is realized by introducing an occupation probability function with an empirical tunable parameter. It can easily be plugged into existing wire-length distribution models without impairing their strengths. Depending on the value of the tunable parameter, the proposed model can give a longer average wire length, as is often desired. The tunable parameter could be interpreted to reflect non-idealities such as anisotropic partitioning strategy and sparsity of cells. The proposed pragmatic approach might possibly serve to save early-stage disappointment at interconnect prediction.