Improvement and Performance Evaluation of an Adaptive Method for Integrated Circuits Pre-silicon Verification

This paper presents an adaptive pre-silicon integrated circuit verification method based on a machine learning algorithm. This approach can overcome the traditional corner-based Process-Voltage-Temperature verification limitations considering variables coverage. The method was intensively tested in what regards its accuracy on a large set of synthetic test functions that mimic the behavior of real circuits. The use of these functions offers the opportunity for a large number of evaluation tests with much lower resource costs compared to the testing of the simulated real circuit. The consistently and the algorithm accuracy is also validated by the results obtained by testing the method on a real Low Dropout Voltage Regulator circuit.