Re: is order urgent doubt

lawrence.jones@xxxxxxxxxxx writes:
Eric Sosman <esosman@xxxxxxxxxxxxxxxxxxxx> wrote [re double double]:

[*] Although one wonders why `long double' wasn't used
instead.

Because double double (using a pair of doubles to represent a single
floating-point number) in the general case doesn't adhere to the
standard's floating-point model and thus isn't a valid implementation of
long double. To be a valid long double implementation, you would have
to restrict the exponents to make the fraction bits contiguous rather
than allowing a gap or overlap.

I don't quite understand. Are you talking about the layout of the
fraction bits in the representation of the type? I don't see how
making then non-contiguous would violate the C standard (assuming
__STDC_IEC_559__ isn't defined).

How is a pair of doubles used to represent a single floating-point
number anyway?

--
Keith Thompson (The_Other_Keith) kst-u@xxxxxxx <http://www.ghoti.net/~kst>
Nokia
"We must do something. This is something. Therefore, we must do this."
-- Antony Jay and Jonathan Lynn, "Yes Minister"
.

Relevant Pages

  • Re: Working with floating point values
    ... That's what I've always done in the past, but decimals are *much* slower to ... complex calculations thousands of times each, ... decimals and doubles is quite noticeable. ... >> nature of floating-point calculations to produce values like ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: is order urgent doubt
    ... Because double double (using a pair of doubles to represent a single ... floating-point number) in the general case doesn't adhere to the ...
    (comp.lang.c)
  • Re: OADate
    ... stores them as standard IEEE 64-bit ... floating-point numbers ". ... I of course meant Doubles! ... The rest of the explanation still holds good though. ...
    (microsoft.public.vb.general.discussion)
  • Re: florating point and VHDL
    ... If you really, really want floating-point hardware, there ... throw away the unused less-significant bits of the fraction, ... Jonathan Bromley, Consultant ... DOULOS - Developing Design Know-how ...
    (comp.lang.vhdl)
  • Re: How to "print" a "double" ?
    ... floating-point value into an integer exponent and a normalized ... fraction and printing them separately. ...
    (comp.unix.programmer)