Dissertation Defense

Functional Error Diagnosis, Correction and Layout Repair of Digital Circuits

Kai-Hui Chang

Given the dramatic increase in design complexity of modern electronics,
digital systems are often released with many latent errors — thereby
demonstrating our inability to ensure their functional correctness. While
improvements in verification allow engineers to more efficiently find
a larger fraction of design errors, little effort has been devoted to
fixing such errors, making it an expensive and challenging task.
In my dissertation we propose innovative data structures, algorithms, and
methodologies that provide a fundamentally new way of solving the error-repair
problem. Our contributions include: (1) Butramin, a bug trace minimizer;
(2) CoRe, a counterexample-guided error-repair technique for gate-level
netlists; (3) REDIR, an automatic RTL error-repair method; (4) InVerS,
an incremental verification methodology for physical synthesis; (5) SymWire,
a symmetry-based rewiring technique based on a new comprehensive
symmetry detector; and (6) a post-silicon error-repair methodology that
automatically fixes electrical and functional errors. To evaluate our
techniques and promote interoperability between verification and debugging,
we propose a framework, called FogClear, that integrates all these components
and matches error-repair requirements at different design stages. By
integrating scalable components in a unified framework, FogClear can greatly
accelerate error correction, improve design quality, and reduce design cost.

Sponsored by

I. Markov, V. Bertacco