[hal-01023862] Stochastic Reliability Evaluation of Sea-of-Tiles Based on Double Gate Controllable-Polarity FETs

Promising devices like Double Gate Field Effect transistors [1] (DGFET) are proposed as basic elements to form tiles for a new kind of nanofabric [2]. For this fabric, regular arrays of tiles are expected in order to build a Sea-of-Tiles. DGGET transistors enable a concise expression of computational function. In order to facilitate this building, a specific logic synthesis methodology based on BBDD was proposed in [3]. We propose in this paper to evaluate the reliability of an application mapped onto this kind of fabric. Based on the BBDD representation of the application, we propose a more accurate stochastic method that takes care of the mapping of the application onto the regular fabric. Stochastic methods for reliability evaluation such as [4] [5] rely on bit streams or probabilistic transfer matrices. Unlike these methods, we propose to model the reliability problem of the nanofabric by means of Bayesian network. The usage of Bayesian methods is not completely new [6] [7] in the context of nanotechnology. Nevertheless, in this paper, we explore Bayesian methodologies for DGFET Sea-of-tiles nanofabric, by integrating them at the application level. This enables us to consider offline reliability strategies as well as online ones based on diagnosis capabilities of Bayesian networks, showing a real new opportunity for EDA tools

from HAL : Dernières publications http://ift.tt/1V4y797