It's not a 'static' analysis, and therefore would p(ossi|roba)bly give you only a subset of the information you seek per trial, but Devel::NYTProf might give you a helpful view.