profiling

[Matthew Fluet]

> I was more worried about the self compile.  That had definitely failed at some
> point in the last few weeks, and I didn't have time to look into it.

Well, the G0 compiling a G1 worked fine, but I don't have access to a
machine to do a G1 to G2 compile.  

> I'm not worried about the time stuff.

Me neither.  I think that as long as you do an "apples-to-apples"
comparision, (i.e., MLton.RUsage.rusage, or even Posix.ProcEnv.times),
then things will work out.