Cell-Aware Model Generation Using Machine Learning
Abstract
Characterizing cell-internal defects of standard cell libraries is an essential step to ensure high test and diagnosis quality. However, such a characterization process, called cell-aware model generation, usually resorts to extensive electrical defect simulations that are costly in terms of run time and utilization of SPICE simulator licenses. Typically, the generation time of cell-aware models for few hundreds of cells may reach up to several months considering a single SPICE license. This chapter presents a methodology that does not use any electrical defect simulation to predict the response of a cell-internal defect once it is injected in a standard cell. More widely, this methodology uses existing cell-aware models (generated from electrical simulations) from various standard cell libraries and technologies to predict cellaware models (learning-based) for new standard cells independently of the technology. Experiments done on several industrial cell libraries using different technologies demonstrate the accuracy and performance of the prediction method.
Fichier principal
Cell-Aware Model Generation by Using Machine Learning - final.pdf (1.25 Mo)
Télécharger le fichier
Origin | Files produced by the author(s) |
---|