Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: How to merge required operations under a common name



On Wed, 27 Jan 2016 14:55:31 +0100, Paul Zimmermann wrote:
>  if we use integers (for example the mpz_t layer of GMP) as
> number format, do we get a valid 1788-2015 implementation?

The basic requirement is that the set of "FP" numbers in a type be
such that each interval result has a unique tightest representation.
So if the underlying format is, say, rationals p/q, the format must
have a well-defined maximum denominator.  You can have a family of
such formats, but each operation would have to specify the result
format explicitly -- otherwise "tightest" would be undefined.

Michel.
---Sent: 2016-01-27 14:03:26 UTC