[clean-list] Re: 1.Floating Point Support (clean-list digest)

[Richard A. O'Keefe]

	Once a teacher asked in a newsgroup how he can iterate the 'e'
	function.

I don't understand what "iterate the 'e' function" means.
"e" is a constant, approximately 2.71828.  I am guessing this
means something like
    x[0] := whatever
    x[i+1] := exp(x[i])...
but what is the point of that?	Or does it mean
"compute e using the formula e = sum from k to infinity of 1/k!"?

	the PowerPC only supports 64bit (the 68040 simulated 96bits or so).
Apple's SANE library (also available for Apple IIs) did support an extended
format, in software as well as in hardware.  SANE was respected as one of the
most faithful implementations of IEEE 754 arithmetic around.

	Apple or Motorola dropped the enhanced (optional) precision.

More accurately, they were unable to persuade IBM to add it to the POWER
architecture, and Apple were unwilling to sink the effort into providing
support for a floating-point model that was dramatically slower than the
other models.

To be completely accurate about it, the IEEE 754 standard envisages *FOUR*
floating point systems:  Single, Single Extended, Double, and Double Extended.
Single Extended was not commonly implemented, on the grounds that Double
could do its job.  Neither of the Extended formats was specified as precisely
as the basic formats.  The intent was that you would use the Extended formats
for accumulating dot products.  The Intel and Motorola systems, although they
differed in the number of bits available, both conformed to the Double
Extended requirements.  But if conforming to the requirements didn't buy you
portability, lots of users didn't want to bother with it.  So there was less
value to vendors in supporting it.

Oh the joys of using a math library with an 80-bit cos() computed using
a 64-bit version of PI (not Apple's)!