Tue Mar 26 11:01:32 EDT 2013

Fast floating point modulo

It seems that the use case of fast float modulo usually comes with
wanting both the integer quotient and the floating point remainder.
The deeper question is whether integers should be reprsented
explicitly, or whether the int modulo should be just the `div' part.

Currently it seems too much work to add integer support.  This is just
an optimization.  It's perfectly fine to work with floats.  So for
now, I'm going to take away the f_mod primitive and change it to