New Methods and Coverage Metrics for Functional Verification

An ever increasing portion of design effort is spent on functional verification. The verification space as the set of possible combinations of a design´s attributes is likely to be very large making it infeasible to verify each point in this space. State-of-the-art verification tools tackle this problem by using directed random generation of combinations in conjunction with manually defined corner cases in order to get satisfactory coverage with the desired distribution. In this work, the underlying methodology to automatically generating complete sets of disjoint coverage models on the basis of formal attribute definitions is extended to take relational constraints into account. This allows the utilization of coverage models with non-orthogonal, non-planar boundaries, which can make whole analysis for coverage data obsolete. It shall be demonstrated, how the proposed methodology can be used to automatically determine corner cases more accurately than it is possible with conventional approaches