Do not use this code!
This is some benchmarking code I wrote several years ago and came back to a few times, in support of a blog post I finally published. One of the reasons I delayed finishing that post so long was that I realized the methodology used here is flawed. It's enough for us to make a (weak) point, but shouldn't be taken seriously.
I'll do something like this again soon, and hopefully that will set a better example.
There is both a Makefile that will do things, and a Python program
(run-experiment.py) that will do similar things. The Makefile is
broken but convenient. Many of the options for programs here don't
actually do anything.
You have been warned.