METHOD AND SYSTEM FOR IDENTIFYING RARE-EVENT FAILURE RATES

摘要

Method and system for estimating defect rates in design. A set of N Monte Carlo samples (points) is derived from a random distribution describing process variations in the design. A subset of these samples is then selected and a subset of the N _init samples is simulated (by a circuit simulator) to measure the performance value for each sample. Then, using a value of N _init process point of a training input, and using the corresponding N _init of performance values as a training output, the model is constructed. Candidate Monte Carlo samples come from N Monte Carlo samples that have not yet been simulated. Each candidate is simulated in the model to obtain a predicted performance value, and the samples are ordered in ascending (or descending) order of the predicted performance values. Then, the simulation of the candidate samples is started in this order. The model and ordering can be updated periodically. Sampling and simulation will be terminated if there is sufficient confidence that all defects have been found.