Accurate and efficient estimation of logic circuits reliability bounds

Walid Ibrahim, Marwa Shousha, John W. Chinneck

    Research output: Contribution to journalArticlepeer-review

    17 Citations (Scopus)

    Abstract

    As the sizes of CMOS devices rapidly scale deep into the nanometer range, the manufacture of nanocircuits will become extremely complex and will inevitably introduce more defects, including more transient faults that appear during operation. For this reason, accurately calculating the reliability of future designs will be extremely critical for nanocircuit designers as they investigate design alternatives to optimize the tradeoffs between area-power-delay and reliability. However, accurate calculation of the reliability of large and highly connected circuits is complex and very time consuming. This paper presents a complete solution for estimating logic circuit reliability bounds with high accuracy in reasonable time, even for very large and complex circuits. The solution combines a novel criticality scoring algorithm to rank the reliability of individual input vectors with a heuristic search to find the input vector having the lowest reliability. The solution scales well with circuit size, and is independent of the interconnect complexity or the logic depth. Extensive computational results show that the speed of our method is orders of magnitude faster than exact solutions provided by Bayesian network exact inferences, while maintaining identical or sufficiently close accuracy.

    Original languageEnglish
    Article number6797873
    Pages (from-to)1217-1229
    Number of pages13
    JournalIEEE Transactions on Computers
    Volume64
    Issue number5
    DOIs
    Publication statusPublished - May 1 2015

    ASJC Scopus subject areas

    • Software
    • Theoretical Computer Science
    • Hardware and Architecture
    • Computational Theory and Mathematics

    Fingerprint

    Dive into the research topics of 'Accurate and efficient estimation of logic circuits reliability bounds'. Together they form a unique fingerprint.

    Cite this