From owner-sc22wg5@open-std.org Tue Feb 28 20:22:30 2006 Return-Path: X-Original-To: sc22wg5-domo2 Delivered-To: sc22wg5-domo2@open-std.org Received: by open-std.org (Postfix, from userid 521) id 810FB1B5D2; Tue, 28 Feb 2006 20:22:30 +0000 (UTC) X-Original-To: sc22wg5@open-std.org Delivered-To: sc22wg5@open-std.org Received: from dkuug.dk (ptah.dkuug.dk [195.215.30.66]) by open-std.org (Postfix) with ESMTP id E03C31B5BC for ; Tue, 28 Feb 2006 20:22:28 +0000 (UTC) Received: from mail1.cray.com (mail1.cray.com [136.162.0.111]) by dkuug.dk (8.12.10/8.9.2) with ESMTP id k1SKM5Zb093625 for ; Tue, 28 Feb 2006 21:22:11 +0100 (CET) (envelope-from longb@cray.com) Received: from mh1750-1.us.cray.com (mh1750-1.us.cray.com [172.31.74.55]) by mail1.cray.com (8.13.3/8.13.3/gw-5323) with ESMTP id k1SJrMZj006536 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Tue, 28 Feb 2006 13:53:22 -0600 (CST) Received: from mh2650-1.us.cray.com (mh2650-1.us.cray.com [172.31.74.50]) by mh1750-1.us.cray.com (8.13.4/8.13.3/hub-5273) with ESMTP id k1SJrMu4022128; Tue, 28 Feb 2006 13:53:22 -0600 Received: from cray.com (mh-dhcp-172-31-20-251.us.cray.com [172.31.20.251]) by mh2650-1.us.cray.com (8.13.4/8.13.3/spool-5907) with ESMTP id k1SJrLKe019360; Tue, 28 Feb 2006 13:53:21 -0600 Message-ID: <4404AC32.6010401@cray.com> Date: Tue, 28 Feb 2006 14:01:54 -0600 From: Bill Long Reply-To: longb@cray.com Organization: Cray Inc. User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.4) Gecko/20030624 Netscape/7.1 X-Accept-Language: en-us, en MIME-Version: 1.0 To: van.snyder@jpl.nasa.gov CC: sc22wg5 Subject: Re: (j3.2005) (SC22WG5.3350) TS for Co-Array Fortran? References: <20060228014031.A6D321B56E@open-std.org> In-Reply-To: <20060228014031.A6D321B56E@open-std.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Cray-VirusStatus: clean X-Spam-Score: 0 () Sender: owner-sc22wg5@open-std.org Precedence: bulk Van Snyder wrote: >It may be desirable to publish a standard for Co-Array Fortran, using >the ISO "Technical Specification" (formerly Type-2 Technical Report) >mechanism. > While the motivations below are well intended, I think that producing a TS would be, overall, a distraction. > >First, it might be published well before the 2008 standard. This would >encourage earlier implementations. > Earlier implementations will be encouraged by customers asking for co-arrays in purchase contracts. This is already starting to happen, even for cluster type systems based on commodity processors. Affected vendors will be able to get a copy of the relevant paper from the j3 archive as easily as, and sooner than, a published version of the same information in the form of a TS. > >Second, it could be posed as a revision of the 2003 standard, allowing >vendors to figure out exactly how it amends Fortran 2003, rather than >hoping they can figure that out from the 2008 standard. This also would >encourage earlier implementations. > The final edits paper for co-arrays will be against the f03 standard. In some areas it may include edits caused by collisions with other features, but those will be noted in the paper. > >Third, if development work on Co-Array Fortran is not largely complete >at the next J3 meeting, it may mean slipping the schedule for Fortran >2008. If it is developed as a TS, and development work is essentially >completed during the integration phase of the 2008 standard, even if the >TS is not published by then, it can be incorporated then, perhaps >without slipping the 2008 standard's schedule. This essentially allows >nearly an extra year for primary development work for Co-Array Fortran. > I very much expect that the edits paper will be done before the next meeting. I would expect the overhead of producing the TS a more likely cause of delaying completion of f08. How we label any work on the feature beyond the next meeting seems like a distinction without a difference. > >Fourth, the current state of development work, together with papers >already written, means the additional work to produce a Technical >Specification as opposed to simply plugging along to integrate it into >Fortran 2008, is nearly complete. > > To the extent that the integration effort results in minor changes, these may be important for the vendors to know about and take into account. The bulk of the implementation work for co-arrays is in implementing remote loads and stores, allocatable co-arrays, and co-array structures with allocatable and pointer components. The syntax and semantics for these parts of the proposal have been stable for some time. Vendors that want an early start on implementing co-arrays can proceed with these parts based on currently available information. By the time most of them get beyond that point we should have made significant progress toward a solid f08 draft. Cheers, Bill -- Bill Long longb@cray.com Fortran Technical Support & voice: 651-605-9024 Bioinformatics Software Development fax: 651-605-9142 Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120