APPLICATION OF MACHINE LEARNING, DATA MINING AND BIG-DATA METHODS IN THE FIELD OF FUNCTIONAL VERIFICATION

Abstract

Summary: Functional Verification (FV) is the process of checking that the design under verification conforms to its functional specification. Functional verification can be done by simulation/emulation using the design under verification with its associated testcases or formally by using static design checkers and formal based techniques. During the execution of the functional verification flow a huge amount of data is generated. And a lot of data exchange between different verification activities is happening. The complexity of many verification tasks can be dramatically reduced if the data generated from one verification step can be used to guide the next verification step and narrow down its work scope. Advanced data analysis techniques can be used within the FV flow to learn knowledge from the data generated out of one step and use it to guide further steps in the verification flow. This thesis demonstrates the leverage of Data Mining, Machine Learning and Big-Data techniques within the state-of-the-art functional verification flows to help optimizing the verification task complexity and effort.