both exhaustive and randomized testing do not make use of the formal semantics, a point that is usually in their favour. In this work, we explore how formal semantics can be used to seed an effective and automatic test generation strategies for instruction sets.