Ada Floating-point model

Ada Floating-point model

[Geert Bosch]

I am trying to understand the floating point model of Ada.
Hopefully somebody would like to answer my questions below.

Does a model interval associated with a value of a floating
point type include one or two model numbers? 

The RM defines a model interval as an interval whose bounds
are model numbers, but doesn't tell which bounds are included
in the interval. 

Then there is a remark that the model interval of a model number
consists of that number only. So when I use two model numbers as
operands of a predefined operation, the result should be in
the smallest model interval containing the result. When this
result is a model number this interval consists of that
number only and the operation is exact, right?


Regards,
   Geert

Re: Ada Floating-point model

[Matthew Heaney]

In article <5sq0qt$aev$1@gonzo.sun3.iaf.nl>, Geert Bosch
<geert@gonzo.sun3.iaf.nl> wrote:

>I am trying to understand the floating point model of Ada.
>Hopefully somebody would like to answer my questions below.

I can't answer you directly, as this is the very stuff I'm still trying to
learn.  But here's at least something.  Ada 83 was influenced by the "Brown
model," documented in

A Simple but Realistic Model of Floating-Point Computation
W. S. Brown
ACM Tranactions on Mathematical Software
Vol. 7, No. 4, December 1981, p 445-480

Even today, after many years of Ada programming, I'm still not sure what a
"safe" number is, though I think I understand what a model number is.

--------------------------------------------------------------------
Matthew Heaney
Software Development Consultant
<mailto:matthew_heaney@acm.org>
(818) 985-1271