Re: Revision 162491 -- fortran -fwhole-file regressions

[Toon Moene <toon at moene dot org>]

Dave Korn wrote:

> On 24/07/2010 07:39, Tobias Burnus wrote:

> > I think reverting the patch is the wrong solution. If you use
> > -ffloat-store, the tests are successful. The reason is that one now
> > inlines the function and uses the result from the register instead from
> > memory. And as the x87 coprocessor has an excess precision, the
> > comparison fails. As on x86-64 one uses SSE it works there.
>
>   Ah, this is Return of the Revenge of PR323 Strikes Back!  :-P
>
>   Could fortran be made to usefully implement -fexcess-precision=standard?
> I'm surprised this hasn't been more of an issue given the prevalence of big
> number-crunching apps in the fortran world.

That is because Fortran (the Language) makes no promises.  All 
arithmetic is "an approximation to the mathematically exact result".

So basically, if you really, really want to compare two REALs (or DOUBLE 
PRECISION numbers) for equality, you have to know what you're doing.

Cheers,

--
Toon Moene - e-mail: toon@moene.org - phone: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
At home: http://moene.org/~toon/; weather: http://moene.org/~hirlam/
Progress of GNU Fortran: http://gcc.gnu.org/gcc-4.5/changes.html#Fortran