x86 update

[Stephen Weeks]

> I think that the new backend.fun created a new bug; the mlyacc benchmark
> diverges in the backend/allocate-registers phase (this time I'm sure it's
> that phase) under the nj-mlton version of MLton (and the
> x86-self-compiled), using both code-generators. 

It doesn't diverge.  It just takes a really long time.  On my 733 MhZ machine,
it took 1410 seconds.  This is extremely bizarre, since the change to
backend.fun was for stuff after register allocation, and so should not have
affected register allocation at all.  And the previous version took about 100s,
IIRC.  I am investigating.

> Note, I think in
> Steve's scripts for running the benchmarks, he special cases the mlyacc
> benchmark (-no-polyvariance?).  

Only the kit is special cased, and it isn't used as a benchmark anymore.