The growing complexity and higher time-to-market pressure make the functional verification of modern large scale hardware\nsystemsmore challenging. These challenges bring the requirement of a high quality testbench that is capable of thoroughly verifying\nthe design. To reveal a bug, the testbench needs to activate it by stimulus, propagate the erroneous behaviors to some checked points,\nand detect it at these checked points by checkers. However, current dominant verification approaches focus only on the activation\naspect using a coverage model which is not qualified and ignore the propagation and detection aspects. Using a new metric, this\npaper qualifies the testbench by mutation analysis technique with the consideration of the quality of the stimulus, the coverage\nmodel, and the checkers. Then the testbench is iteratively refined according to the qualification feedback. We have conducted\nexperiments on two designs of different scales to demonstrate the effectiveness of the proposed method in improving the quality\nof the testbench.
Loading....