A Probabilistic Context-Free Grammar Based Random Test Program Generation

The aim of this paper is to show the use of a probabilistic context-free grammar in the domain of stimulus generation, especially random test program generation for processors. Nowadays, the randomly constructed test stimuli are largely applied in functional verification to verify the proper design and final implementation of systems. Context-free grammar cannot be used by itself in this case, because conditions for instructions of the program are changing during the generation. Therefore, there is a need to introduce additional logic in the form of constraints. Constraints guarantee the continuous changes of probabilities in the grammar and their application in order to preserve the validity of the program. The use of the grammar system provides a formal description of the stimuli, while the connection with constraints allows for the wide use in various systems. Experiments demonstrate that this approach is competitive with a conventional approach.