CDC floating (was: Re: It All Comes Down to the IBM 704)

In article <IJDCCv.86@xxxxxx> "*** T. Winter" <***.Winter@xxxxxx> writes:
>In article <1120916134.777467.284420@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>
jsavard@xxxxxxxxx writes:
> > *** T. Winter wrote:
>...
> > > I have no idea in what sense the exponent can be viewed as being in
> > > excess-n notation. See the following table (the exponent field is
> > > given in octal):
> > > field exponent
> > > 3777 infinite
> > > 3776 1022
> > > 2001 1
> > > 2000 0
> > > 1777 indefinite
> > > 1776 -1
> > > 0001 -1022
> > > 0000 -1023
>...
> > I thought the table was something like this:

> > 2001400000000000 1
> > 2000400000000000 0.5
> > 1777400000000000 0.25
>
>I think you should reread. 0.25 is 177640000000000. When the exponent
>field is 1777 the f.p. number is undefined (a special class).

Yes, all 1777 xxxx xxxx xxxx xxxx were indefinite, including
1777 4000 0000 0000 0000, in CDC Cyber floatin point.
The normalized value of 1.0 was 1720 4000 0000 0000 0000,
with mantissa 2^47 and exponent -57 octal = -47 decimal.
The normalized value of 0.25 was 1716 4000 0000 0000 0000.
--
If we don't protect our forests, the bushlands will become Bushlands.

pmontgom@xxxxxx Microsoft Research and CWI Home: Bellevue, WA
.