This is the mail archive of the fortran@gcc.gnu.org mailing list for the GNU Fortran project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Possible explaination why large_real_kind_form_io_2.f90 is failing on PPC/OSX


On Sun, Jan 14, 2007 at 11:30:13PM +0100, Dominique Dhumieres wrote:
> > If you already have a populated obj/, then "make" should only
> > rebuild simplify.o, and "make install" will do the installation.
> > This should only take 5 minutes.
> 
> Sorry, but I am building gfortran through a fink script which handles the
> patches and the install for me. The drawback is does remove the temporary
> directory at the end (I don't think it is able to restart from it anyhow).
> So I have to redo the full build at least of gfortran and it takes sometime
> (hours). Yes I know I'll have to learn a less barbaric way, but my knowledge
> of unix tools is only marginally better than my knowledge of C (say 10*0!-).

Oh my that is ugly.

> BTW what should I expect from the change? I guess something will be written
> on stderr, but what?

The intermediate value of nearest(x,y) that is held in an mpfr type.
You showed a constant-folded value of nearest(1._16,1._16) = 1.0e0.
I expected a long string of digits, and with a precision of 31 digits
I suspect that the 32nd digit (if printed) is greater that 5.  The
middle or back end may be rounding the value back to 1.e0.

-- 
Steve


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]