Deriving Code Coverage Information from Profiling Data Recorded for a Trace-based Just-in-time Compiler
Code coverage information is typically recorded by adding instrumentation to an application before or during execution. However, this has the disadvantage that the instrumentation decreases the performance.
A high-performance virtual machine (VM), such as Oracle's Java HotSpot VM, is already designed to record profiling data for the executed application. This profiling data is intended to guide optimizations during just-in-time (JIT) compilation. In this paper, we propose a runtime system that allows deriving exact code coverage information from this recorded profiling data by guaranteeing certain system properties. With this approach, we avoid the hassle of instrumenting the application explicitly. Therefore, we minimize both the run-time overhead and the implementation effort for recording code coverage information.
We implemented our approach for a variant of the Java HotSpot VM that uses a trace-based JIT compiler. However, our approach is simple and general enough to be applicable to most modern VMs. Our evaluation with industry-strength benchmark suites shows that this approach allows recording code coverage information while retaining almost full peak performance. So, our approach is also suitable for collecting code coverage information in day-to-day operation environments such as long running server applications.