extract_floating() vs floatformat_to_doublest()

[Andrew Cagney]

> Hello,
> 
> Now that I've isolated the doublest code (doublest.[hc]), the next step is to rationalize the number of interfaces available for doing floating point conversion.  Currently there are:
> 
>     floatformat_to_doublest()
>     extract_floating()
> and
>     floatformat_from_doublest()
>     store_floating()
> 
> I've two possabilities in mind:
> 
>     o    dump extract/store floating
>         (fixing floatformat_*_doublest())
> 
>     o    dump floatformat_*_doublest
>         (possibly replace *_floating()
>         with extract,store_floatformat())
> 
> my current preference is the former.  Does anyone have a compelling argument for the latter?


To make this more concrete (and hopefully easier to understand :-) see 
the attatched.  It implements part of the former.  It also illustrates 
how everything should be routed through the host conversion routines 
before falling back to the software FP conversion code.

	Andrew