Re: motion elementary functions
On pow(xx,yy): I agree with Hossam Fahmy and Dan Zuras that arguments
based on the rationality of floating-point representations (and the
concomitant dependency on the base) are irrelevant in a real-REAL-based
interval context -- except for the following:
> You see the problem is that when x < 0 the function xµy
> is defined at only countably many discrete points. In
> a floating-point domain with radix 2 or 10 there are
> none for r = 2 & only finitely many for r = 10.
Dan forgot plain old small integers (where the base does not matter).
The pown and rootn functions only make sense for point second arguments
(i.e. individual integers, not intervals).
On function names (and name clashes): IEEE 754 is language-independent
and only defines what must or should be there, and how what is there
must behave -- not what it is called. After all, languages have lexical
rules that may conflict with any given prescribed set of names. So name
clashes are easy to avoid, as they only apply to the *symbolic* names
used within a standard.
I believe the same should hold for 1788 -- and we should indeed use the
same names as in 754. A separate column that mentions corresponding
function names in a few popular current languages or libraries could
be a useful non-normative (i.e. purely informative) addition.
What the standard SHOULD say is that conformance requires documenting
the mapping between names defined in a language or library and the
corresponding P1788 (or IEEE 754) standard names.
Michel.
---Sent: 2009-10-14 10:31:59 UTC