contification paper

[Stephen Weeks]

> Why don't you do them both the same, whatever you decide.  I'm saying that
> those two benchmarks compiled and executed on a different machine
> anyways, it makes sense to have them the same, even if we use
> -no-polyvariance for them.

Actually, I'm going to make executables and let you run them on your machine.
That way, the only flaky numbers we have will be the compile times.

I think I'll choose -no-polyvariance.