Dear Fortran programmers,
The US and international Fortran committees (J3 and WG5) have been
working diligently on a new revision of Fortran scheduled to be
published in 2010. The main feature in this revision, currently still
being drafted, is intrinsic explicit parallelism based on Co-Array
Fortran (Numrich & Reid). This extension of the current Fortran 2003
language is similar to the extensions to C made in UPC and to Java in
Titanium.
To help you in learning more about co-arrays there are two WG5 white
papers, one an overview of the features, by John Reid, found at:
ftp://ftp.nag.co.uk/sc22wg5/N1651-N1700/N1697.pdf
and the other a Rationale for some of the technical decisions made in
the design of the base language, by myself, found at:
ftp://ftp.nag.co.uk/sc22wg5/N1701-N1750/N1702.pdf
I would like to ask for comments on the proposed co-array features from
all of you. Compiler developers, experts in parallel programming and
languages, and especially users of Fortran, are welcome to comment. The
most useful comments are technical in nature and focused, so that they
can be taken into account with reasonable amount of work and
modifications to the existing specifications. Both praise and critique
are welcome.
I personally think that explicit parallelism in the major
programming languages for scientific computing is essential for
increasing the productivity of scientific programmers and believe that
co-arrays are an essential addition to Fortran 2008. However,
there is a danger that co-arrays will not be included as a mandatory
part of Fortran due to internal divisions in the committee. I would
prefer if the final decision is influenced by the community.
You can submit comments (or questions) to myself via e-mail (and I will
relay them accordingly), or directly to WG5's convenor John Reid
([log in to unmask]). Official papers can also be submitted to J3 (US
committee) for technical consideration and development at:
http://j3-fortran.org/upload.html
Note that the next WG5 meeting is in Las Vegas in February.
Please share these with others that you think are interested. The
earlier we get comments, the better we can respond. Once an FCD official
draft of the standard is prepared, it will be important to submit formal
comments to your country representative body, but more on that later.
Thank you,
Aleksandar
--
Aleksandar Donev, Ph.D.
Lawrence Postdoctoral Fellow @ LLNL
High Performance Computational Materials Science and Chemistry
E-mail: [log in to unmask]
Phone: (925) 424-6816 Fax: (925) 423-0785
Address: P.O.Box 808, L-367, Livermore, CA 94551-9900
Web: http://cherrypit.princeton.edu/donev
|