Dynamic Verification. Dynamic Verification is also known as test or experimentation. This type of testing involves the execution of a system or its components [20]. There are numerous methods available; as an example for embedded software test methods, the book Testing Embedded Software [21] gives a good overview.
Dynamic Verification. Dynamic verification is of a simpler nature than static verification. Essentially it consists of building meaningful input scenarios, simulate them, make sure they have touched a certain functionality of the design, and check that the output was as expected. Due to its simplicity, simulation-based dynamic verification is the de facto industry verification method. However, simulation struggles with achieving completeness. There are 'clever' ways to mitigate the issue and a few of them are mentioned in Section 3. It all boils down to the quality of the simulation being performed, which is a hard characteristic to quantify. Multiple testbenches can be built in order to achieve a high quality verification. The term testbench has been used to describe varied solutions, ranging from a very basic set of pin-level stimulus run standalone in a simulator, all the way to a major part of a sophisticated verification environment containing complex models and interfaces. These environments and their terminologies change according to what methodologies are being adopted, but in general they can be broken down into three parts: input generation, checking schemes, and monitoring assets. Higher quality verification can be reached by increasing the meaningfulness of the inputs created by input generators, as well as by increasing the quality of the sequences of inputs being generated. This difficulty of building meaningful inputs is known as the generation problem. There are approaches that try to automatically guide the generation process [17], but in general this task relies on the expertise of the verification engineer, his knowledge of the design, and his knowledge of the current coverage holes that remain to be verified. Dynamic verification can make use of a golden model when one is available, to which the design can be compared against. Having models that behave like entire SoCs is borderline infeasible but models of smaller portions of the system can be used for selective checking. The standard approach when using a model is to submit the same inputs generated by a testbench to both the design and the model. When a mismatch is found, the testbench should be able to detect that automatically. This process is referred as self-checking, and it is a fundamental part of a verification environment. Simply put, a checker has to be able to say if a certain testbench was a pass or a fail. On top of that, when a mismatch is detected, a (good) checker ideally has to try to pinp...
