Functional verification is used to confirm that the logic of a design meets its specification. The most commonly used method for verifying complex designs is simulation-based verification. The quality of simulation-based verification is based on the quality and diversity of the tests that are simulated. However, it is time consuming and compute intensive on account of the fact that a large volume of tests must be simulated to exhaustively exercise the design functionality in order to find and fix logic bugs. A common measure of success of this exercise is in the form of a metric known as functional coverage. Coverage is typically indicated as a percentage of functionality covered by the test suite. This thesis proposes a novel methodology to construct a model using SVM, Gradient Boosting Classifier and Neural Networks aimed at replacing random test generation for speeding up coverage collection.
