| Dan Nagle 2008-02-18, 7:22 pm |
| Hello,
I suppose I should summarize what happened at 183,
even if just to keep rumurs to a minimum.
There has developed some opposition to coarrays
on WG5, due to the size of the feature overall, and to experience
with other parallelization schemes that didn't work as planned (HPF).
So some want a cautious, go-slow approach.
The result is that coarrays were divided into a core part,
which remains in the draft standard. This core is basically
what has been implemented by Cray now. There is also
a further coarrays part, which was moved to a TR (to be standardized
in a later revision of the standard). This consists of items
which were added to the basic coarrays feature in the expectation
that they would be useful, but that are not implemented by any compiler now.
Thus, we'll get to tinker some more with the details.
Thus, the overall size of coarrays in f08 is reduced. But most
of the coarrays features have been retained. The TEAM I/O was
removed, at least for now.
The Fortran standard is committed to coarrays.
In the draft, coarrays are spelled without the hyphen.
The schedule, as far as I can see now, will have a Public Comment
period during June of this year. The Committee Draft of the standard
will likely be available for review by March. So everyone will have
time to read it carefully before commenting. The Editor has the draft
nearly ready now, but the posting process will take a few days. J3
will reply to comments at its August meeting.
The process can be improved by greater participation by all concerned.
Contact your National Body (J3 in the USA) with your views.
There are some photos of meeting 183 on J3's web site.
I've given a longer summary of 183 to Ian Chivers (who attended)
for publication in Fortran Forum. (Ian is in some of the photos.)
On 2008-02-18 06:18:31 -0500, Simulate <li.simula@gmail.com> said:
> http://www.nag.co.uk/sc22wg5/
--
Cheers!
Dan Nagle
|