x86 Update

[Stephen Weeks]

> Here are the latest performance numbers for x86-G1 MLton.  

x86-G1!  I'm curious to hear what the speedups for the self-compile are.

> benchmark                        executable-size
>                                C      x86  x86/C   
...
> fib                        33223    32559   0.98
...
> tak                        33247    32655   0.98

These are about 10K too large.  Probably just due to some dead code elimination
not yet in your snapshot.

> All in all, I think a successful conclusion.

I'll say.  Faster on every benchmark (checksum is noise :-P -- anyways, we're
already 8x faster than NJ on that one).  I like the compile times too, but I
still bet there's even more room there.
	
> Here's what's on my short-term/mid-term todo list:
...

Your priorities sound good to me.  Maybe we should all sit in on a summer-end
conference call next week to do a little planning and decision making.  Matthew,
what does your time table look like?  When do you think would be a good time to
merge in your stuff?  Would you like me to try do so with the snapshot that you
made today?