From: dblock <dblockdotorg@...>
Date: 2011-01-19T14:54:09+09:00
Subject: Re: Improving performance of hash math

Here're some actual numbers. I am doing the above code 10K times with
a data distribution that's somewhat not-random (some overlap in keys).

 %self     total     self     wait    child    calls  name
 37.61      1.49     0.99     0.00     0.50    61446  Hash#each
 10.00      0.26     0.26     0.00     0.00   419946  Float#+