On Sun, 3 Nov 2002, der Mouse wrote:

> > The F_float is a 32-bits.  It's sign magnitude number with a sign
> > bit, 7-bits of exponent, and 24-bits for the fractional part.  The
> > range is given as 0.29x10^-38 through 1.7x10^38.
>
> 7 bits of exponent gives you a range of only 1.08e-19 through 9.22e18
> (2^-63 to 2^63); you need 8 bits of exponent to get the range of
> 0.29e-38 to 1.7e38.
>
> I believe it's 1 bit sign, 8 bits exponent, 24 bits mantissa - but that
> 24 bits includes the hidden bit, which is not actually stored.

Oops, you're correct.  Here are some ASCII versions of what's in
the VARM:

F_float
                     15
31                 16|14      7 6      0
 +------------------+-+--------+-------+
 |    fraction      |S|  exp   | fract | :A
 +------------------+-+--------+-------+


D_float
                     15
31                 16|14      7 6       0
 +------------------+-+--------+-------+
 |    fraction      |S|  exp   | fract | :A
 +------------------+-+--------+-------+
 |    fraction      |    fraction      | :A+4
 +------------------+------------------+
63                                     32


G_float
                     15
31                 16|14        4 3    0
 +------------------+-+----------+-----+
 |    fraction      |S| exponent |fract| :A
 +------------------+-+----------+-----+
 |    fraction      |    fraction      | :A+4
 +------------------+------------------+
63                                     32


H_float
                     15
31                 16|14               0
 +------------------+-+----------------+
 |    fraction      |S|  exponent      | :A
 +------------------+-+----------------+
 |    fraction      |    fraction      | :A+4
 +------------------+------------------+
 |    fraction      |    fraction      | :A+8
 +------------------+------------------+
 |    fraction      |    fraction      | :A+12
 +------------------+------------------+
127                                    96


-brian.