| Entering the nanometer era, a major challenge to current design methodologies and tools is how to effectively address the high defect densities projected for emerging nanotechnologies. To this end, in this dissertation we propose a reconfiguration-based defect-tolerant design paradigm for defect-prone nanoelectronic technologies. In our paradigm, designs are mapped into a nanofabric comprised of reconfigurable regions, architected using a suitable hierarchy of design abstractions, so as to meet the target yield with best expected performance. The new design goal is thus to devise an appropriate structural/behavioral decomposition which improves scalability by constraining the defect mapping and reconfiguration process to small fabric regions, while meeting a desired probability of successful instantiation, i.e., yield.; A key feature of our proposed nanofabric architecture is that it enables the defect mapping and configuration tasks to be performed within the nanofabric itself, eliminating the costly per-chip offline processing. Specifically, we have devised a novel group testing method that can systematically identify defective components and/or connectivity in a fabric region. It enables the entire fabric to be tested and configured in a scalable way, using a relatively small number of easily configured triple-modular-redundancy (TMR) test tiles executing concurrently on different regions of the target nanofabric.; Moreover, our proposed design paradigm offers a rich framework in which critical trade-offs among performance, yield, and complexity can be explored. The probabilistic nature of these tradeoffs has required us to introduce a new class of 'reliability-aware' high-level synthesis (HLS) problems. In particular, rather than carefully optimizing a single ('deterministic') solution, as done in traditional HLS, our approach requires the joint synthesis and optimization of a sufficiently large family of alternative solutions, so as to achieve the specified target yield, with best-expected performance. We have developed a Reliability-Aware Synthesis framework for NANOfabrics (RAS-NANO), aimed at systematically solving this new class of 'reliability-aware' HLS problem. It enables designers to effectively explore the complex probabilistic design space associated with the new reconfiguration-based defect-tolerant design paradigm. |
