Fortran 2003/2008 features and compiler poll - real(kind=16) support

[Jerry DeLisle]

Tobias Burnus wrote:
> I now announced that webpage on lang.comp.fortran. But feel also free to 
> vote yourself.
> 
> http://physik.fu-berlin.de/~tburnus/F2003-polling/
> 

Tobias,

I had one user of gfortran talk to me in person on this subject.  For normal day 
to day work he uses pgf. Recently however, due to some circumstances gfortran 
was needed and "saved the day".  (I suspect a hardware problem)

Regardless, the one feature needed most was real(kind=16) capability on x86 
hardware. One of the drivers for this is the need to test sensitivity of 
computations to round-off error.

I wonder if we should not get this in our sights.  FX and I have discussed this 
in the past and we know gcc has existing 128 bit float libraries that can be 
used.  The trick is getting it working with gfortran.  There is probably a lot 
of configury magic to do.

I need comments from others please and is anyone interested in working on this?

Best regards,

Jerry