Referring to lexical variables from within a Benchmark is the mistake made most often when benchmarking. And a very large percentage of the benchmarks I've seen in the past years make this mistake. It has prompted me to give a talk about this (and other mistakes) at YAPC::NA last year.

Moving my %h from BorisZ code to the main program is wrong. Moving my %bucket from my code to the main program is wrong (although for a different reason!).

Abigail