> depending on the exact usage, a large Java program can be slightly slower, as fast, or a little faster than a comparable C++ codebase

This is probably true but…

Does anyone here have any data to back that up?