Unifying i387 support

[Eli Zaretskii]

> > Does anyone care for the MMX view of the ST(i) registers?  Perhaps, if
> > we design a change in the layout, it would be better to make it
> > last...
> 
> I don't understand.  They do appear last.

No, I meant to *add* the MMX view to the structure, as if they were a
separate set of registers.  Then GDB could display either the FP view
or the MMX view, depending on the mode that the processor is in, or
perhaps we could have a sub-command to tell GDB what view to use.

(FWIW, ``to make it last'' was meant to say that the new layout should
support newly-introduced features, so that it won't become obsolete
too quickly.  Sorry for unclear wording.)

> > Also, why did you choose to represent the registers as char buffers
> > instead of what they are (shorts, longs, long doubles, etc.)?
> 
> Because I'd like to support cross-debugging from architectures with
> different word sizes and FP formats.

I thought the structure you presented was the platform-independent
one, and that therefore all the conversions required by architecture
differences have been done already by the platform-dependent code that
fills the structure.