Making Programs Faster

The Uncertainty Principle

There's another way to try to eliminate bias
- You can try to correct for it
- By adding a lot of complicated machinery to measure bias and subtract it from the results
This is the Benchmark.pm approach
But if it goes wrong, you have no idea what really happened

      null: -1 wallclock secs
        (-0.07 usr +  0.01 sys = -0.06 CPU) 
        @ -16666666.67/s (n=1000000)

Even when it goes right, you have no idea what really happened

"There are two ways of constructing a software design: One way is to make it so simple that there are obviously no deficiencies and the other way is to make it so complicated that there are no obvious deficiencies."

-- C. A. R. Hoare

These days I always write my benchmarks manually
Or I have Benchmark::Accurate write the script for me