[MLton] Shortening dependency chains in the x86 codegen

[Florian Weimer]

* Vesa Karvonen:

> Quoting Florian Weimer <fw at deneb.enyo.de>:
>> > Take a look at the following code for 32-bit endianness swapping
>> > generated by MLton on x86:
>> 
>> I'm sure that nowaydays, you could just use bswapl.  It's available in
>> i486 and up, and AFAIK, call compatibles in use today implement it.
>
> Just to clarify, I wasn't suggesting the optimization for the sake
> of optimizing endianness swapping.

Ah, okay.  Have you actually measured that this is a problem?
(I'm curious how you'd discover such code generation issues.)