This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: gfortran precision
On Sun, 2004-08-08 at 17:06, Paul Brook wrote:
> On Monday 09 August 2004 00:57, Jerry DeLisle wrote:
> > On Sun, 2004-08-08 at 16:40, Paul Brook wrote:
> > > On Monday 09 August 2004 00:34, Jerry DeLisle wrote:
> > > > What is the highest floating point precision supported by gfortran on a
> > > > IA-32 architecture (ie P4)? Is it possible to get to KIND=16?
> > >
> > > gfortran currently only supports IEEE single (4-byte) and double (8-byte)
> > > floating point formats.
> > >
> > > Note that the P4 doesn't support quad precision in hardware. It does
> > > support an 80-bit "extended" format, which is then padded to 12 or 16
> > > bytes.
> > >
> > > Paul
> >
> > Are there plans to support quad precision or the 80-bit extended
> > format? I am assisting some coworkers in getting setup to use gfortran
> > and one has a specific application that needs the higher precision.
>
> I/We plan to support any floating point format gcc supports on a target. For
> i386 means single, double and 80-bit extended formats.
>
> I can't give any estimate on when this might be implemented though. Things
> get done when someone volunteers or pays for the work to be done.
>
> Any formats not supported by the gcc backend (eg. quad precision on i386) will
> not be supported by gfortran. In theory it would be possible to add emulated
> quad-precision support to the i386 backend, but that's a somewhat different
> proposition.
>
> Paul
Is there something I could do as a volunteer. I am not an expert on
compiler technology, but I might be able to do janitor work.
Jerry