floats and intervals, constants, 0.1, 1e400 and Inf
Baker Kearfott wrote, with respect to constant->interval conversion:
> By "constant" do you mean a 754 floating point number
> or an external character representation?
Both. Note that there are several different 754 types, and we have to
cover the case where the source type does not match the target type,
e.g. the type used to represent the interval bounds.
> Standardizing midpoint-radius arithmetic would complicate 1788's task
> significantly. However, when we have our formal procedures in place,
> someone can propose it and we can vote on whether we want to do it.
Interval Arithmetic lends itself to abstract definition better than plain
floating-point arithmetic, so it is conceivable that the standard would
have two components, the first of which is independent of the actual
representation. This first part would acknowlege that there may be
different interval types, possibly in the same program. In the latter
case conversion rules between types have to be defined.
Some representation types have unique issues, such as representation of
semi-unbounded intervals in midpoint-radius form. That may be a sufficient
reason to stick to lower-upper-bound representations, but one still has to
cover the existence of multiple types for the bounds themselves.
Michel.
Sent: 2008-11-10 16:17:42 UTC