ISO/IEC JTC/SC22/WG5 N1759 Minutes of Meeting of ISO/IEC JTC1/SC22/WG5 Hosted by IPSJ/ITSCJ, the Japanese National Member Body for JTC 1 in Tokyo, Japan November 16-21, 2008 List of Attendees: John Reid (JKR Associates, UK) Convenor Dan Nagle (GMU, USA) PL22.3 chair Ian Chivers (Rhymney Consulting, UK) Malcolm Cohen (PL22.3, NAG, UK) Masakazu Hayashi (Fujitsu Limited, Japan) Ichiro Homma (Hitachi Ltd., Japan) Rex Jaeschke (SC22 Chair) Michael Ingrassia (PL22.3, Sun Microsystems Inc., USA) Steve Lionel (PL22.3, Intel Inc., USA) Bill Long (PL22.3, Cray Inc., USA) Toon Moene (PL22.3, Gnu Fortran, Netherlands) Nick MacLaren (University of Cambridge, UK) Steve Morgan (University of Liverpool, UK) Secretary David Muxworthy (BSI, UK) Chair of Drafting Committee Kazuo Nishimura(Komazawa University, Japan) Fumitoshi Sato (University of Tokyo, Japan) Hiroyuki Sato (University of Tokyo, Japan) Jane Sleightholme (Fortranplus, UK) Van Snyder (PL22.3, Caltech/JPL, USA) Masayuki Takata (Edogawa University, Japan) Minoru Tanaka (Fujitsu Limited, Japan) Jim Xia (PL22.3, IBM, Canada) Note: INCITS/PL22.3 is abbreviated throughout to PL22.3 and its papers are referenced as 08-nnn. They are available from http://www.j3-fortran.org/ WG5 papers are referenced as Nnnnn. They are available from http://www.nag.co.uk/SC22WG5/ 1. Opening of the Meeting The meeting opened at 0910, Sunday 16th Nov. 2008. 2. Opening Business 2.1 Introductory Remarks from the Convenor John Reid welcomed everyone to the meeting and thanked the Japanese delegation for hosting the meeting. He explained that the meeting would be a mixture of WG5 (with John Reid in the chair) and PL22.3 (with Dan Nagle in the chair) business. 2.2 Welcome from the host Masayuki Takata welcomed everyone to Tokyo. 2.3 Local arrangements. Several trips had been arranged to local places of interest. 2.4 Appointments for this meeting The drafting committee will consist of David Muxworthy, Toon Moene, Van Snyder, Masayuki Takata, and Jim Xia with David Muxworthy as chairman. Steve Morgan would act as secretary. John Reid would act as librarian. 2.5 Adoption of the agenda (N1732) The agenda was adopted as published. 2.6 Approval of the minutes of the Las Vegas Meeting (N1715) The minutes were approved. 3. Matters arising from the minutes There were no matters arising not covered by the agenda. 4. Status of the Las Vegas Resolutions (N1714) The proposed TR on Enhanced Coarray features was mentioned. No SC22 approval had been requested as yet. 5. Reports 5.1 SC22 Matters N1742 (summary of important items at the SC22 meeting) had been posted by John Reid. He reported that there had been a discussion on submission of comments in SC22 ballots. It had been agreed that where comments were extensive, summaries could be submitted in the standard template with references to further papers. These additional papers would become SC22 documents and would be required to be submitted at the same time as the ballot. SC22 had expressed concern about the importance of preserving the history of programming language standardization. Van Snyder offered to pursue the possibility of scanning some of the committee's paper records. John Reid had been re-appointed as WG5 Convenor for a further three year term. 5.2 National Activity Reports (Heads of delegations) US - very little to report other than PL22.3 business Netherlands - no report Canada - no report Japan - N1746 UK - N1757 5.3 Report from Primary development body (PL22.3 Chair) Dan Nagle noted there had been more comments on the CD than for Fortran 2003. Responses had been given to all comments. 5.4 Reports from Other Development Bodies It was noted that PL22.3 Standing document 18 (cumulative edits for the technical corrigendum) was not yet updated. 5.5 Liaison Reports C - Dan Nagle reported that the C committee is producing a document that is concerned with standardising thread programming (Linux threads, Posix threads, and Win threads). MPI - Dan Nagle noted that some additions to the Interoperability TR had been motivated by liaison with the MPI committee. N1755 gives details. UPC - there had been a successful workshop at George Washington University. UPC committee are going for an international standard. PL22.3 IFIP/WG2.5 - Van Snyder went to Toronto in August. There will be an effort to standardise Interval Arithmetic within IEEE. OpenMP - no report OWGV(now SC22/WG23) - Dan Nagle reported that there would be language- specific clauses in the document. Fortran would probably be the first to be completed. Dan mentioned that WG23 was getting strong support from the US DoD, US DHS, and UK MoD. SC22 (Monday) John welcomed Rex Jaeschke (SC22 Chair) who was attending the meeting. Rex introduced himself and summarised his role as SC22 chair. He noted that SC22 had now agreed that teleconferencing could be used for meetings. He outlined the broad rules under which it could operate for subgroups. He stressed that this was not intended as a replacement for face- to-face meetings but could be used to carry out business where it was deemed appropriate and agreed by the different parties. 6. Consider the ballot comments on the draft revision (N1723) of Fortran 2003 and decide on any technical changes to be made. The UK votes on the CD were considered first. David Muxworthy went briefly through the comments. (N1740) There was nothing to add to what had been said at the last meeting. There was then a discussion on the Parameterised Derived Types (PDT) feature which had caused implementers of Fortran 2003 some difficulty. One vendor had now implemented them successfully and another was close. Four other vendors were not currently implementing them. One vendor was concentrating on the object-oriented feature. There was some discussion about a possible Corrigendum that might remove those parts of PDTs that were giving vendors special difficulty. Only complete removal of LEN parameters was suggested. There was no consensus and the approach was eventually dropped as a way forward. The Japanese delegates were supportive of the addition of MPI features. It was agreed that close liaison with the MPI committee was desirable and Mike Ingrassia offered to act as liaison. John Reid summarised the discussion by noting that clearly Fortran needed to provide better support for MPI. The meeting then moved to a technical discussion of the comments on coarrays in N1744 and N1745 (UK comments). The main issue discussed was that of whether VOLATILE coarrays should be allowed in the language. There seemed to be general agreement that the functionality enabled by having VOLATILE coarrays was needed (e.g. spin-loops) but that there may be better/safer ways of providing at least some of the functionality. This could be provided by language features and/or intrinsic procedures. Following some broad discussion the issue was deferred to subgroup. There was some discussion on whether coarrays should be included as a separate part of the standard. Most of this discussion repeated arguments put forward at the last meeting. Eventually John Reid concluded that no real consensus could be reached and that, since a majority of countries wanted coarrays in Part 1, this was the only way forward. There was some discussion of N1748 (Memory model for coarrays). The N1743 proposal (G format extension) was deferred to subgroup. Canada's comments were then considered. Some outstanding issues (finalisation of coindexed objects (see 08-199) and co-indexed objects in type bound procedures) were deferred to subgroup. PL22.3 paper 08-162 (disallow allocatable array dummy argument corresponding to coarray actual argument) had been approved by PL22.3. There was then some discussion on locks (see 08-256r2) which was eventually deferred to subgroup. The German comments were then considered. The issue of changes to the Bessel functions was deferred to subgroup. The US comments (N1741) were then considered. The editorial and technical repairs were agreed, except for 08-199r1, 08-207, 08-214, and 08-216r2, which were deemed to be feature requests. All the feature requests were deferred to subgroups. ............................................................................... Monday It was agreed to take no action on the German comments concerning Bessel functions. A reply to the comments was in 08-211r3. ............................................................................... Tuesday The issue in paper 08-293r1 (a new COMPLEX intrinsic) was discussed. It was suggested that it should include all the functionality of CMPLX. Also suggested was the syntax of a COMPLEX constructor (like a derived type). Straw Vote ---------- COMPLEX as in 293r1 2 COMPLEX intrinsic 6 Don't do either 5 Undecided 4 Straw Vote ---------- Should we have an intrinsic function 8 Should we just use a syntax not restricted to constants 3 Undecided 5 Shall we do it as a function Y 3 N 6 U 6 Since there was no consensus WG5 decided to drop the COMPLEX intrinsic feature. Paper 08-296 (G Format) was considered. Straw vote ---------- In favour of something like 296? Y 10 N 2 U 5 Happy with 296? Y 10 N 0 U 7 There was some discussion of 08-295 sections 2 & 3 (External Procedures and Binding Labels). John Reid requested comments. There were none forthcoming and so it was assumed that everyone was happy with the edits. John Reid created draft documents N1760-n (n=1,2,...) to summarise the responses to all the comments on the CD. This document would evolve to N1760 during the meeting. ............................................................................... Wednesday John Reid announced that he had contacted Wolfgang Walter who had indicated that Germany (although not at the meeting) was in favour of having coarrays in Part 1 of the standard. This was an indication of how they would vote on the 2008 standard and thus reinforced the decision to continue with coarrays in Part 1. Paper 08-300(LOCK feature) had been discussed in PL22.3 mode and John Reid briefly consulted with the meeting to verify that WG5 concurred with its contents. Paper 08-297 (atomic memory operations) was discussed. A US caucus had agreed on its contents. It was suggested provision should be made for the possibility of more than one kind of atomic logical and more than one kind of atomic integer. Straw Vote: Is current proposal sufficient? Y - 10 N - 1 U - 4 WG5 approved the approach. There was a discussion of 08-290r1 (Replies to coarray and volatile issues) Straw Vote: Should we leave in the edit requiring coarray volatility to be consistent? Y - 10, N - 0, U - 5 There followed a UK caucus and Dave Muxworthy reported back that the UK were as happy as could be expected that the UK comment on VOLATILE had been addressed. N1760 (response to comments) was discussed and it was agreed that all its contents should be agreed at the WG5 level. ............................................................................... Thursday 08-301 (Referencing procedures using % and coindexing) discussion. Straw Vote: Individual Country Do 143r3 3 Canada, Japan Do 301 12 USA, Netherlands, UK Do Nothing 0 0 Undecided 1 0 It was agreed in principle to go with 08-301. WG5 then agreed that integrating STORAGE_SIZE (as in 08-302, US TAG item) was desirable (there were no objections). There was some discussion on optional dummy arguments in BIND(C) interfaces. Also new features requested by MPI (assumed type and assumed rank). There was some further discussion of the draft N1760 (response to comments) and Dave Muxworthy noted that only part of UK comment had been addressed thus far. The schedule as in N1693 was discussed. Several members (vendors and users) could not see the benefit of publishing the standard as early as specified in the current schedule (as in N1693). On the other hand at least one vendor would have F2003 by end of this year and would have F2008 by the end of 2009. Another would have one by mid-2009. Some vendors could see customer demand for 2003 and 2008 whilst there were some who clearly did not and would not have F2003 in the foreseeable future. Concerns expressed were: there was no experience of F2003; features such as LOCK were being introduced very late. Following a lengthy discussion there was no agreement on changing the schedule (a 6 month delay had been mooted) and it was decided to defer a decision to the next meeting in May 2009. ............................................................................... Friday Discussion continued on 08-303 (request for TYPE_IS_EXTENSIBLE). 08-147r1 (issues with VALUE dummy arguments). Straw Vote: Individual Country Agree with 147r1 Y - 7 US, Netherlands, UK N - 2 Canada U - 7 Japan It was decided to go with 147r1. 7. Construct a preliminary draft of the TR on further interoperability with C Discussed 08-305 (Extended Interoperability TR-29113). Bill Long was thanked for producing this draft. The main issue was over optional arguments. Straw Vote: Should we have optional arguments? Individual vote: Y - 7, N - 2, U - 5 Country vote : Netherlands - U, US - Y, UK - Y, Japan - Y, Canada - N John Reid stated that this should be regarded as direction - it was not a final decision. At this point the discussion was stopped due to lack of time. There would be a letter ballot of WG5 on the TR as a whole and separately on the optional arguments issue. 8. Consider the Fortran defects reports (interpretations) in 006 PL22.3 worked on interpretations during the meeting. 9. WG5 Business & Strategic Plans 9.1 Goals for 2008-2011 The primary goal for 2008-2011 is the completion and publication of Fortran 2008. The schedule for Fortran 2008 will be reviewed at the meeting in May 2009. Work on interpretations will continue. TRs on further interoperability with C and with further coarray features will be constructed. 10. Closing Business 10.1 Future meetings 4-8 May 2009 - Joint meeting in Las Vegas, Monday 0800 to Friday 1200. Feb 2010 - Las Vegas(tentative). Aug 2011 - London, Copenhagen, Canada and Germany were possibilities. 10.2 Any other business John Reid expressed warm thanks on behalf of WG5 to the Japanese delegation for hosting the meeting and providing excellent facilities and social programme. 11. Adoption of Resolutions (N1758) T1-T6 were accepted unanimously (individual vote and country votes). T7 was passed 15-0-1. All countries voted yes except Canada who abstained (4-0-1). T8-T10 were accepted unanimously (individual vote and country votes). T11 was accepted unanimously (individual vote and country votes). T12 was accepted unanimously (individual vote and country votes). 12. Adjournment The meeting adjourned at 1200 (Friday 21st November, 2008)