Evaluating Workloads Using Multi-Comparative Functional Coverage

In this paper we present a technique for comparing multiple tests and workloads. We show how to automatically determine what each test does uniquely and how to present the information as succinctly as possible. This technology has a number of uses, including in understanding the contribution of tests in regression buckets, especially of legacy systems, and in evaluating what is missing in tests compared to customer usage. We also show that the technology used in the analysis is superior to previous technology in that it can automatically find holes that were previously only found manually.