Automation and Optimization of Coverage-driven Verification

This paper proposes a new method for automation and optimization of coverage-driven verification (CDV) of hardware systems that is based on evolutionary computing. In comparison with the standard CDV that utilizes random search, using this method, the convergence to the maximum coverage is much faster, fewer input stimuli are used and no manual effort is required from the user. Moreover, the optimization is targeted to the verification process itself without the dependence on the circuit that is verified. For the demonstration purposes, the sequential arithmetic-logic unit (ALU) and the RISC processor were verified and the results show that using the optimization method, a significant improvement can be achieved.