I apologize if I gave the impression I was claiming otherwise. It's just context for the discussion.
When Veedrac dismissively tells you - "pidigits tests whether you have bindings to GMP" - you should ask why he hasn't told you that the measurements can be different even when all the programs use GMP; you should ask why he hasn't told you that the measurements also show the difference for the same language implementation when programs do and don't use GMP.
etc etc