Use double's instead of floats

[kay]

There should be support for doubles. Floats lose precision very quickly. If a person is timing something in nanosecond units then after 17ish milliseconds precision will be lost by your library. If they expect to be making nanosecond measuements over values over 20 milliseconds regularly, then their statistics will be wrong.

Float for this reason often isn't used in mathematical operations unless memory or computational sympathy are an issue.