large compile times for mlyacc

[Stephen Weeks]

> Unfortunately, rolling back didn't fix the problem, although I thought for
> sure I had run self compiles with the codegen using the loopForest stuff.

By rolling back (directed-graph.sml), I assume you mean to 1.21?  So
if that didn't speed stuff up, it must have been the introduction of
1.21 that caused the slowdown, right?

> Choices are:
>  1. somebody look at loopForest and figure out how to speed up nest
>  2. forget about using loopForest info in codegen
>  3. (stopgap) the codegen should work o.k. if x86LoopInfo.createLoopInfo
>      returns information that makes it look like no block is part of a
>      loop (the codegen only uses to weight stuff carried in registers
>      and to insert alignment at loop headers).
> 
> I can probably cobble 3 together in a little while.

(3) seems fine for now (it is essentially a temporary version of (2)).
Someday I hope to get around to reading that loop headers paper and
looking at your code.  But it's pretty low on the todo.