[Proj] Re: Re: Proj4 Bug (rtodms)

[Gerald I. Evenden]

On Thursday 09 November 2006 4:44 pm, Glynn Clements wrote:
	...
> > I am still suspicious of the setting of the FPU's rounding flag.
> >
> > This does raise in my mind: what should the default state of the rounding
> > flag be?
>
> Appendix F of the C99 standard requires round-to-nearest during
> translation (F.7.2) and at program startup (F.7.3). However, the
> standard doesn't require an implementation to conform to appendix F
> unless it indicates conformance by defining the macro
> __STDC_IEC_559__.

I gave up getting the standards years ago as they were getting to damn 
expensive so I rely on Harbison & Steele, v.5.  But I am glad to see that 
there is a (tentative) default setting.

> In any case, 240.0/60.0 should be exactly equal to 4.0 regardless of
> the rounding mode; anything else is a bug. The rounding mode is only
> relevant if a result cannot be represented exactly.

All this would be irrelevant if we all got 64 bit machines and could do this 
in integer arithmetic.
 
> Contrary to widespread belief, the FPU doesn't get the least
> significant bits of a result from /dev/random ;)

Oh?  Damn!  I learn something every day.  :-)

I really did learn something: I didn't know there was a /dev/random
-- 
The whole religious complexion of the modern world is due
to the absence from Jerusalem of a lunatic asylum.
-- Havelock Ellis (1859-1939)  British psychologist