Efficient coverage analysis metric for HDL design validation

Abstract

Simulation is still the primary approach for the functional verification of register-transfer level circuit descriptions written in hardware description language (HDL). The major problem of the simulation approach is to choose a good metric to gauge the quality of the test patterns. The finite state machine (FSM) coverage test can find most of the design errors in a FSM. However, it is impractical for large designs because of the state explosion problem. In the paper, a higher-level FSM model is proposed to replace the conventional FSM model in the coverage test. The state transition graph can be significantly reduced in the model so that the complexity of the test sets becomes acceptable, even for large designs. This higher-level FSM model, called the semantic finite state machine (SFSM) model, can be easily extracted from the original HDL code automatically with little computation overhead. The advantages of using this model instead of the conventional FSM model in HDL design validation are thoroughly discussed. The implementation results show that it is indeed a promising functional coverage metric.