Yes exacly. And I assume that the test suite of all of those project are long enough to average the usual jitter of wall time mesurement.

What I'm hoping to see is if rust+llvm vs rustc+gcc binary speed are within a few percents or if there is a real difference between the two (I'm expected that we eventually reach the former once thinLTO and other optimisations are implemented).

And while doing so it could also be possible to measure the difference in max RSS.