My understanding is that MATH_EMULATE is used in the kernel config solely
to determine whether the kernel will use FP instructions (and hence
registers).  By turning on MATH_EMULATE, you cause kernel FP operations to
be emulated, so the kernel doesn't need to fret with the FP registers,
and/or if the kernel has the FP registers saved, it may then freely use
them as temporaries for data-copy loops.

(Not that my understanding necessaarily has any bearing on reality.  But
that's how I've understood the situation for a few years, now.)


My understanding is that this relates to the question answered in one of
the FAQ's:

  http://www.netbsd.org/Documentation/kernel/#why_msoft_float

(Hm...I cannot see where -msoft-float is ever specified...)


I'm not too sure of my understanding, but if I'm wrong, then it's time
that I found out.  (^&


  "I probably don't know what I'm talking about." --rauch@eecs.ukans.edu