On Wed, 2006-11-29 at 08:44 +0100, Birger Heede wrote:
> I think Mike's writeup is a very good intro:
> http://www2.hursley.ibm.com/decimal/
It is indeed, with a treasure trove of references to boot. Thanks for
the pointer.
I've wondered sometimes why we never saw hardware support for rational
numbers. Probably the kicker is the requirement for arbitrary-length
integers? Too bad.
But I'd guess there's no reason you couldn't implement floating-point
"registers" as rational numbers. The writeup referenced above says in
part:
"The working precision of the arithmetic is not
necessarily determined by the representation,
but may be freely selectable within the limits
of the representation as required for the problem
being solved. Implementations may provide very
high precision if they wish."
So the "registers" could be implemented under the covers as rationals,
and then externalized in DFP representation when a store comes along.
It would be nice to set a condition code at that time to let the unwary
know whether a precision error occurred during the store.
--
David Andrews
A. Duda and Sons, Inc.
[EMAIL PROTECTED]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
