Constraint Propagation Methods for Robust IC Design

Konferenz: ZuE 2015 - 8. GMM/ITG/GI-Fachtagung Zuverlässigkeit und Entwurf – Reliability by Design
21.09.2015 - 23.09.2015 in Siegen, Deutschland

Tagungsband: GMM-Fb. 83: ZuE 2015

Seiten: 8Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Autoren:
Krinke, Andreas; Lienig, Jens (Dresden University of Technology, IFTE, Dresden, Germany)
Jerke, Goeran (Robert Bosch GmbH, Automotive Electronics, Reutlingen, Germany)

Inhalt:
Constraint engineering is one of the key enabling technologies to address robustness and reliability issues in today’s IC designs. Design constraints are used to express and verify the customer’s demands and the designers’ intent. These constraints put limits on some design object’s parameter values and, hereby, represent additional design information that is then used to enforce robustness, reliability and other design targets. In hierarchical IC designs, the states of constraints often depend on parameters of other design modules in the design hierarchy. Therefore, these constraints are propagated within the design hierarchy to be considered when taking design decisions and performing design verification. This propagation is an essential component of any robustness-, reliability- and constraint-driven design flow. To the best of our knowledge, this is the first work that presents a systematic classification and detailed discussion of the constraint propagation problem. Despite the vast number of conceivable constraint types, we found that there are only six different propagation categories. We derived a single but generic constraint propagation algorithm for all six propagation categories. Our work closes a critical automation gap in today’s constraint engineering flows by proving the full automatability of the constraint propagation problem and by providing a comprehensive and consistent propagation solution. We also present experimental results from an industrial design that demonstrate the applicability for large design problems.