On 2004.4.2, at 04:06  PM, J. MacPhail wrote:
>
> Hypothesis: the actual floating point instructions only appear in a
> very few files (such as the main math library, "libm" or whatever it
> is called).

As Bruce pointed out, those "few files" are used in too many places.

Many years ago, UNIX was mostly a clot of integer functions. These 
days, encryption, hashing, and many other not-obviously-numeric 
functions are borrowing floating point functions right and left.

My personal armpits is that it might be worth a separate port, say 
macint68k or something, at least for a release or two.

Wish I had time to back my opinions up with work, but I've been trying 
for more than five years to break out the resources to do what Bruce 
has done. Maybe I let my wife have her way too much.