(j3.2006) A loss for Fortran

[Toon Moene]

On 03/17/2017 07:47 PM, Rafik Zurob wrote:

> Using a C++ outer layer and a Fortran computation core is quite common
> nowadays.  I'm interested though in how this related to coarrays.  ECMWF
> experimented in 2014 (via the CRESTA project) with using coarrays to
> improve communication performance (
> http://www.ecmwf.int/sites/default/files/elibrary/2014/13673-challenges-getting-ecmwfs-weather-forecast-model-ifs-exascale.pdf
> ).  If it's not confidential, may I ask whether coarrays are now used in
> production code and whether this change has any effect on that?  I
> remember when I interacted with them in the past that they were using
> several Fortran 2003 features.  Are they going to continue to improve the
> Fortran portions of the code and would they be adopting newer Fortran
> standards or staying at the current level they have?

I don't think it's confidential if I reply to your mail as the Dutch 
representative in the ECMWF's Technical Advisory Committee (to its Council):

This was a one-off test.

ECMWF's position on *any* new Fortran feature has been: there have to be 
multiple compilers to support that feature (for rather obvious reasons 
that I do not have to repeat on this mailing list).

So they (or others working on the common code) regularly sent out 
messages like:

1. Is the following code conforming to Standard 20XY ?

2. Do you know any compiler (... list ...) that doesn't compile the 
following code:

    <code>

[ using the feature outlined in 1. ]

So the answer about coarrays is simple: As soon as enough compiler 
support them.

Hope this helps.

Kind regards,

-- 
Toon Moene - e-mail: toon at 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/wiki/GFortran#news