[idunning]

The problem is that you may have to stray outside that vectorized bubble, and at that point you enter into a two-language situation if you are using R (e.g. RCpp) or Python (Cython?) - which was one of the motivations for the whole Julia language, to avoid needing to do that.

[epistasis]

But that's no excuse for a bad benchmark. If it actually were necessary to stray out of vectorization, then the Julia benchmark would have a point. But those cases are few and far between in what I have experienced, and apparently in what the benchmark authors have experienced.