Date   : Thu, 01 May 2008 10:50:40 +0100
From   : jgh@... (Jonathan Graham Harston)
Subject: Converting numeric representations

Angus Duggan wrote:
> Jonathan Graham Harston wrote:
> > As normalising is moving the mantissa until the top bit is
> > nonzero, the word denormalising seems appropriate for moving the
> > mantissa (usually the other way) until the exponent is zero.
> 
> No, denormalised numbers are already used in FP. They refer to numbers where
> the exponent is at minimum (a negative number rather than zero), and the
> mantissa doesn't have the leftmost bit set. They are numbers smaller than
> the range which the full precision of the mantissa can be applied.
 
That clarifies things. I shall comment the code as denormalising
the (normalised) floating-point number until the exponent is zero,
{ie, not fully denormalised} at which point, if there was no
under/overflow, the mantissa is the integer version of the real.
 
-- 
J.G.Harston - jgh@...                - mdfs.net/User/JGH
United Kingdom Tercentenary : 1707-2007 : http://yr300.org.uk