Adaptive multi-domain thermal modeling and analysis for integrated circuit synthesis and design

  • Authors:
  • Yonghong Yang;Changyun Zhu;Zhenyu (Peter) Gu;Li Shang;Robert P. Dick

  • Affiliations:
  • Queen's University, Kingston, ON, Canada;Queen's University, Kingston, ON, Canada;Northwestern University, Evanston, IL;Queen's University, Kingston, ON, Canada;Northwestern University, Evanston, IL

  • Venue:
  • Proceedings of the 2006 IEEE/ACM international conference on Computer-aided design
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Chip-package thermal analysis is necessary for the design and synthesis of reliable, high-performance, low-power, compact integrated circuits (ICs). Many methods of IC thermal analysis suffer performance or accuracy problems that prevent use in IC synthesis and hinder use in architectural design. This article describes ISAC, a novel, fast, accurate thermal analysis system for use in IC synthesis and design. We present new, cooperative, temporal and spatial adaptation methods to dramatically accelerate accurate analysis. The proposed system unifies steady-state, time-domain, and frequency-domain analysis techniques. It is composed of our spatially-adaptive multigrid iterative solver, a new temporally and spatially adaptive asynchronous time marching solver, and a new spatially-adaptive frequency-domain moment matching solver. Together, these cooperative adaptation and multi-domain analysis techniques allow the proposed system to efficiently solve the static, short time scale, and long time scale variants of the IC thermal analysis problem. Experimental results demonstrate significant performance improvement over existing thermal analysis solutions. Our spatial adaptation techniques bring a 21.6x -690.0x speedup over recently-published steady-state thermal analysis techniques. Our unified spatial and temporal adaptation techniques, within our asynchronous time marching method, bring a 1,071x -1,890x speedup over other widely-used, time-domain thermal analysis techniques with less than 0.5% error. Our spatial adaptation techniques enable the efficient use of our frequency-domain thermal analysis technique, which brings a 10x -100x speedup over the fastest-known time-domain technique, when used for long time scale thermal analysis. The thermal analysis system described in this article has been implemented as a C/C++ library that has been publicly released for free academic and personal use.