(j3.2006) the IEEE modules and distinguishing zero values

[Malcolm Cohen]

On Wed, 2011-11-16 at 00:53 -0800, Malcolm Cohen wrote:
> (no-one cares about ATAN2)

Van replied:
>except mathematicians who want to be careful.
>
>Read "Branch Cuts for Complex Elementary Functions

I don't need to read it again.  Even mathematicians being careful don't always 
agree on this one, particularly since there are virtually no accuracy or 
rounding requirements on ATAN2 for any other values!

Fred Tydeman wrote:
>In looking at N1830.pdf (ISO/IEC 1539-1:2010), it is
>not clear to me if
>
>  (-0.0) ** (-1)
>  (-0.0) ** (-1.0)
>
>are required, allowed, or prohibited from being -infinity
>for IEEE-754 (IEC 60559) implementations.

That is because ** is not an IEEE operation, and there was/is not consensus on 
making any requirements on the processor in this area.  If/when the IEEE modules 
get extended to handle IEC 60559:2010 it might be a different story, but doing 
that extension will probably require a lot of work since IEC 60559:2010 is much 
bigger and more complicated than the 1989 version.  There seems to be a lack of 
enthusiasm for this at the moment...

Cheers,
-- 
................................Malcolm Cohen, Nihon NAG, Tokyo.