I don't get it. Your benchmark is completely bogus, you admit it in an update, but you only put that in a little remark at the bottom. So you're actually still pretending that everything is just fine.

You should fix the benchmark, for example by assigning the sort result to a lexical array in the subs, and rerun the benchmark, and strike through the flawed benchmark.

But no, you prefer to pretend everything is fine and dandy. You get a downvote from me. What a waste of time.

Update As per my request, friedo updated his node a little, pointing out the benchmark is flawed, there where it got inserted. At least, now people are prewarned. Me happy.