Re: gfortran.dg/large_real_kind_1.f90

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Tue, Sep 27, 2005 at 06:55:15PM -0700, Steve Kargl wrote:
> On Wed, Sep 28, 2005 at 12:23:09AM +0930, Alan Modra wrote:
> > On Tue, Sep 27, 2005 at 10:06:28AM +0930, Alan Modra wrote:
> > > The real problem is that huge() is actually returning infinity.  I'll
> > > see if I can figure out why that is so.
> > 
> > This fixes the value returned by huge() for systems that use the IBM
> > extended double format (or the mips variant) for long double.  It uses
> > a similar hack to the one I made for c-cppbuiltin.c.
> > See http://gcc.gnu.org/ml/gcc-patches/2004-03/msg00388.html.
> > Bootstrap and regression test powerpc64-linux in progress.
> > 
> 
> Your patch appears to break the definition of the Fortran
> model numbers as given in 13.7.1.  In particular,  if we
> have

In hindsight, you didn't break the definition of the model
numbers.  You simply have introduced an inconsistency between
the parameters that define the model numbers and the
value returned by huge().

> program a
>    integer, parameter :: knd = 4 ! Set to IBM Extended kind to see problem.
>    integer b, e, p
>    real(knd) x,y
>    b = radix(x)
>    e = maxexponent(x)
>    p = digits(x)
>    x = huge(x)
>    print*, x
>    print*, (1._knd - real(b,knd)**(-p)) * real(b,knd)**(e-1) * b
> end program a
> 
> The printed values need to be the same.

I forgot that the print statements won't work without FX's
patch.  You may be able to inspect the values via -fdump-tree-original.

-- 
Steve