Efficient online computation of statement coverage

Evaluation of statement coverage is the problem of identifying the statements of a program that execute in one or more runs of a program. The traditional approach for statement coverage tools is to use static code instrumentation. In this paper we present a new approach to dynamically insert and remove instrumentation code to reduce the runtime overhead of statement coverage measurement. We also explore the use of dominator tree information to reduce the number of instrumentation points needed. Our experiments show that our approach reduces runtime overhead by 38-90% compared with purecov, a commercial statement coverage tool. Our tool is fully automated and available for download from the Internet.