numeric_limits

[Oleg Krivosheev]

Hi, Edward

i don't understand how you're going to handle
specialization for float/double, particulary
is_iec559 member and NAN stuff.

It's highly platform dependant and numeric_limits should be 
actually generated on the fly (at least in part) 
in configure/Makefile.

Another interesting question is how numeric_limits
specializarion for float/double will live
together with -ffast-math option.

just a thought...

regards

OK