It would be really good if the author actually linked to somewhere where you can reproduce the benchmark. Either way, I'm pretty sure the correct response is "u

It would be really good if the author actually linked to somewhere where you can reproduce the benchmark. Either way, I'm pretty sure the correct response is "use pypy", not that I can check it