ISO/IEC JTC/SC22/WG5 N2020 Minutes of Meeting of ISO/IEC JTC1/SC22/WG5 Hosted by INCITS/PL22.3, the US Member Body for JTC1/SC22 in Las Vegas, NV, USA June 23-27, 2014 List of Participants: John Reid (JKR Associates, UK) convenor Dan Nagle (NCAR, USA) PL22.3 chair Reinhold Bader (Leibniz Supercomputing Centre, Germany) Daniel Chen (IBM, Canada) Malcolm Cohen (NAG, UK) Bob Corbett (Oracle, USA) Alessandro Fanfarillo (University of Rome, Italy & NCAR, USA) (Monday and Tuesday) Steve Lionel (Intel, USA) Bill Long (Cray, USA) Nick Maclaren (University of Cambridge, UK) Jeanne Martin (former WG5 convenor, USA) Lorri Menard (Intel, USA) Toon Moene (Gnu Fortran, Netherlands) David Muxworthy (British Standards Institution, UK) Craig Rasmussen (University of Oregon, USA) Damian Rouson (Sourcery Inc, USA) (Tuesday and Wednesday) Van Snyder (Caltech/JPL, USA) Masayuki Takata (Edogawa University, Japan) Notes: 1. This was a joint meeting of SC22/WG5 and INCITS/PL22.3. These minutes record only the parts of the meeting in WG5 mode. 2. WG5 papers are referenced as Nnnnn. They are available from http://www.nag.co.uk/SC22WG5/ 3. INCITS/PL22.3 is abbreviated throughout to PL22.3 and its papers are referenced as J3/14-nnn. They are available from http://www.j3-fortran.org/ 4. During the meeting TS 18508 "Additional Parallel Features in Fortran" was referenced by its informal name "Further Coarray TS" and is so recorded in these minutes. 1. Opening of the Meeting The meeting opened at 08:08 on 23rd June 2014. 2. Opening business 2.1 Introductory remarks from the Convenor The convenor said that the principal objectives of the meeting were to develop the draft TS on further coarray facilities and to choose the list of deficiencies and discrepancies in Fortran 2008 to be addressed in the next revision. There would be initial set of straw votes on items that had been proposed and they would then be passed to subgroups for more detailed consideration. 2.2 Welcome from the Host Steve Lionel welcomed participants on behalf of PL22.3. 2.3 Local arrangements Steve Lionel invited participants to dinner at Battista's restaurant on the Wednesday evening. 2.4 Appointments for this meeting The drafting committee was Reinhold Bader, Daniel Chen, Toon Moene, David Muxworthy (chair), Steve Lionel and Masayuki Takata. David Muxworthy acted as secretary and John Reid as librarian. 2.5 Adoption of the agenda (N2010) The preliminary agenda was adopted. 2.6 Approval of the minutes of the Delft 2013 Meeting (N1977) The minutes of the 2013 meeting had been discussed and approved by email during July 2013. It was noted that minutes are now circulated for approval by WG5 letter ballot approximately one month after each meeting. 3. Matters arising from the minutes There were no items not otherwise on the agenda. 4. Status of Delft 2013 Resolutions [N1978] There was nothing to report. 5. Reports 5.1 SC22 Matters (Convener) Convenorship. John Reid's term as convenor would end in September 2014. He was willing to continue for one more three-year term but would stand aside if any other member wished to volunteer for the post. He was encouraged to continue for another three years. Change in ISO rules. Changes to rules meant that every working group member was acting as an individual and did not represent a member body. However since SC22 ballots were conducted on a country by country basis, it was proposed to continue WG5 procedures as before, that is recording country positions as well as taking votes by individual members. National activity reports would be optional. Document archive. ISO required that all working group documents be held on their website. It had been agreed that all SC22 documents should have open access. John Reid was in the process of moving the WG5 archive to the ISO website, rearranging the documents into year folders, rather than using the current system of folders of 50 documents. ISO also required working groups to use their email system. It was hoped that the current email system would continue. WG23 (Vulnerabilities). This matter was postponed to the Vulnerabilities liaison report (5.5 below). 5.2 National Activity Reports Canada: There was little standards activity related to Fortran. IBM had recently released a compiler with more of Fortran2008. Germany: No report. Japan: Preparation for the Japanese version of Fortran 2008 continued but would need more than one year to complete. Japan had been proposing in the SC22 ballot to vote against disbandment of WG23. Netherlands: The Programming Languages Committee of the Netherlands meets twice a year, and parallels SC22 activities. Four language experts are members of the committee, for the following languages: C, C++, Cobol and Fortran. The committee votes, when requested by ISO, on language specific issues. As an example, the vote on Corrigendum 3 of the Fortran 2008 Standard was approval in the March 2014 meeting, after the major problems addressed in the Corrigendum and the work that had gone into it to fix them had been explained. Aside from this regular work, the proposed disbanding of WG23 led to quite some discussion on the mailing list. UK: The report is in N2017. The main item of interest was arrangements for the 2015 WG5 meeting in London. There was also related information in N2018 and N2019. US: The main activity was that of PL22.3. Steve Lionel had been on an INCITS training course. 5.3 Report from Primary Development Body (INCITS/PL22.3) Development had been proceeding to schedule. All items specified by WG5, other than conformance to ISO/IEC/IEEE 60559:2011, had been implemented. 5.4 Reports from other Development Bodies (Editors/Heads) Interoperability TS: Bill Long Development of the Further Coarray TS was proceeding. 5.5 Liaison Reports: NCITS/PL22.11 (C): Dan Nagle The CPLEX work, to provide parallel facilities in C, continues. OpenMP would not incorporate all of OpenACC. MPI: Bill Long There was progress in updating interfacing from Fortran 95 to Fortran 2008 + TS29113. In particular there was progress on modules. It had been agreed to drop Fortran 77 support. A paper "Implementing the MPI-3.0 Fortran 2008 Binding" had been accepted for presentation at the EuroMPI/ASIA 2014 conference in Kyoto on September 10-12, 2014. MPICH 3.1.1 release has Fortran 2008 support. Tests for Fortran 90 have been ported to Fortran 2008 and are working well. There are three relevant Working groups for MPI 4.0: Fault Tolerance (resiliency), Hybrid (interaction with OpenMP, CAF, OpenCL,...), Fortran (yet better interfaces). UPC: Damian Rouson No report. IFIP/WG2.5: Van Snyder An IEEE project would be set up after the July 2014 meeting in Vienna. There was interest in interval arithmetic. John Reid had offered to talk on the Further Coarray TS. OpenMP: Bill Long The OpenMP 4.0 standard would be fully supported in gfortran 4.10, due in April 2015. WG23 (Vulnerabilities): Dan Nagle WG23 had not met since September 2013 as no meeting had been called. A motion in PL22 to propose that WG23 be disbanded had failed but a subsequent PL22 letter ballot had succeeded. The motion had been forwarded to SC22 for an SC22 letter ballot. This was initially accepted but then cancelled on the grounds that such an important decision should not be made by letter ballot. The matter was to be discussed at the SC22 plenary meeting in September 2014. Ada contributors had stopped work until the politics had been resolved. A group of people active in other SC22 working groups but not previously active in WG23 were thought to be the principal motivators in proposing disbandment. WG5 had been asked if it would publish the next edition of TR24772 but WG5 had neither the resources nor the expertise. As well as the Vulnerabilities TR the working group had the project DIS17960, Code Signing for Source Code. 6. Construct the PDTS on Further Coarray Features. Apart from a WG5 straw vote taken on Thursday, development of the PDTS was conducted in PL22.3 mode throughout the week. In total 16 PL22.3 papers with edits to N2007 were approved (viz. 14-141r2, 14-147r1, 14-148r2, 14-151r3, 14-152r2, 14-156r3, 14-157r2, 14-158r3, 14-160r4, 14-172, 14-173r2, 14-185r1, 14-205r1, 14-210, 14-212r1, 14-213r1). 7. Consider the Fortran defect reports (interpretations) in J3/14-006. No defect reports were presented for WG5 consideration at this meeting. 8. Construct list of deficiencies and discrepancies in Fortran 2008 to be addressed in the next revision. To guide discussion at the meeting, straw votes were held on proposals submitted for possible changes to Fortran 2008. N2015 DIN-A SELECT RANK block construct: 3 yes-8 no-4 undecided N2015 DIN-B Extension to SELECT TYPE construct: 2-10-3 N2015 DIN-C Placement allocation: 2-10-3 N2016 UK-11 Assumed coshape: 5-2-8 N2016 UK-12 File-unit connection: 8-2-5 N2016 UK-13 I/O error conditions: 9-1-5 N2016 UK-14 Minimal exponent width: 7-1-6 N2016 UK-15 GENERIC statement as alternative to interface block: 12-0-3 N2016 UK-16 INTENT(IN) and VALUE: 14-0-0 N2016 UK-17 Recursion by default: 7-4-3 N2016 UK-18 Consistency of intrinsic arguments: 14-0-0 N2016 UK-19 COSHAPE intrinsic: 11-0-3 N2016 UK-20 Bitwise logical functions: 7-4-3 N2016 UK-21 Numeric conversion check: 3-4-7 N2016 UK-22 RANDOM_NUMBER intrinsic: The vote was taken on whether this matter needed attention: 13-1-0 US-18 (J3/14-101r1) Add optional type-spec for the ac-do-variable in an array constructor implied-DO: 8-0-7 US-19 (J3/14-128r)1 Making WHERE obsolescent: 1-9-5 US-20 (J3/14-165) PROTECTED attribute for types: 1-6-8 It was decided that edits relevant to the Further Coarray TS agreed during the meeting should be written relative to the TS rather than to the revised language document. [PL22.3 plenary and subgroup sessions from 11:15 to 17:00] The meeting adjourned at 17:00. ........................................................................... Tuesday An early draft of N2023 listed preliminary decisions on whether or not to progress the items for language revision listed above. This document was agreed. As the schedules in N1979 had not been adhered to, an early draft of N2024 was presented which gave three options for scheduling the two principal projects. After discussion they were voted on separately as follows. Plan A: Delay both the Further Coarray TS and the language revision for one year. Straw vote: 10 yes - 2 no - 3 undecided Plan B: Incorporate the further coarray facilities into the language revision which would cause a delay on the language standard of eight months. Straw vote: 5 yes - 7 no - 3 undecided Plan C: Keep the current schedule for the language revision and address the TS to the revision, delaying the TS by two years and eight months. Straw vote: 5 yes - 7 no - 3 undecided [PL22.3 plenary session from 08:40 to 10:30] There were further straw votes on items for which the J3 subgroup /jor had recommended no action. These were: N2016 UK-20 Bitwise logical functions: 10 not do - 3 do - 2 undecided N2016 UK-21 Numeric conversion check: 2 not do - 9 do - 4 undecided There was also a straw vote on a new J3 paper: J3/14-201 undeprecate FORALL: 1 do - 12 not do - 3 undecided [PL22.3 plenary session from 10:40 to 11:45] There was a straw vote during discussion in J3 mode of item N2016 UK-17 Recursion by default: Should elemental functions be allowed to be recursive? 12 yes - 1 no - 2 undecided Alessandro Fanfarillo gave a presentation on "Coarrays in GNU Fortran" in which he compared performance of benchmark coarray programs on various systems. The slides are in document N2022. This was followed by discussion. [PL22.3 subgroup sessions from 12:45 to 16:00] Malcolm Cohen gave a presentation on the changes necessary to make the next revision of Fortran compliant with ISO/IEC/IEEE 60559:2011 (requirements UK-09 and US-14, ref N1982). Details are in papers J3/14-196 and J3/14-198. [PL22.3 plenary sessions from 16:45 to 17:10] Following a brief discussion there was a straw vote on the preferred length of WG5 meetings. The result was: 5.5 to 6 days: 5 - 4.5 to 5 days: 5 - undecided 6. The meeting adjourned at 17:15. ........................................................................... Wednesday [PL22.3 plenary session from 08:00 to 09:45 with WG5 intermissions] Arising from discussions on papers in J3 mode, there were WG5 straw votes on the principle, rather than the specific detail, of adding items to the list of approved proposals. These were as follows. UK-17 Recursion by default: 11 yes - 1 no - 3 undecided UK-11 Assumed coshape: 1 yes - 2 no - 12 undecided this was treated as a 'no' J3/14-202 in and : accepted unanimously J3/14-196&198 (requirements UK-09 and US-14) Conformance to ISO/IEC/IEEE 60559:2011: 9 yes - 0 no - 5 undecided UK-08 Reduction intrinsic REDUCE. Remove REDUCE from N2014: 7 yes - 3 no - 6 undecided. Nevertheless there were two WG5 straw votes on details of REDUCE: Specify a default for a zero-sized array when IDENTITY is absent: 0 yes - 9 no - 5 undecided What is the result of reducing a zero-sized array: 11 error - 0 processor dependent - 5 undecided Subsequent processing in both WG5 (straw votes on Friday) and PL22.3 modes restored REDUCE as an active item. Future meetings were discussed, as recorded at 11.1 below. [PL22.3 plenary and subgroup and US TAG sessions from 10:15 to 16:30] A revised draft of N2024 was discussed and withdrawn for further development. A hybrid of plans A and B was suggested: The TS to be delayed by a year. The revision also to be delayed by a year but incorporation of TS to start on time. Attention was drawn to the first set of draft resolutions on the website. [PL22.3 plenary session from 16:45 to 17:00] The meeting adjourned at 17:00 and was followed by the meeting social event at Battista's restaurant. ........................................................................... Thursday A revised draft of N2024 was considered. This included the new plan AB. In discussion it emerged that plan AB was the most preferred and this was confirmed by a straw vote: 10 yes - 3 no - 2 undecided There was a straw vote on whether to adopt a new proposal to enhance the CMPLX intrinsic function: 8 yes - 0 no - 7 undecided [PL22.3 plenary and subgroup sessions from 08:25 to 17:00 with WG5 intermissions] There was a straw vote on whether the further coarray features should require asynchronous progress: 8 yes - 1 no - 6 undecided The draft resolutions were reviewed and minor changes made. The meeting adjourned at 17:00. ........................................................................... Friday The latest draft of the resolutions was discussed and minor edits suggested. [PL22.3 plenary session from 08:20 to 10:20 with WG5 intermission] There was a straw vote on whether the REDUCE function should have an argument to enforce strict left to right evaluation. This went: 8 yes - 3 no - 4 undecided A further straw vote on whether to allow control of commutability went: 1 yes - 10 no - 5 undecided 9. Review of the current draft of the next revision. The draft was still being developed. 10. WG5 Business and Strategic Plans 10.1 Goals for 2014-2017 These had been discussed throughout the week and led to the plans in N2024. 11. Closing Business 11.1 Future meetings Arrangements for the 2016 WG5 meeting in Boulder were discussed. The straw vote on first preferences for dates was: June 5-10: 10 - July 31-August 5: 3 - August 7-12: 1. On hotels, low price was considered more important than distance from the meeting venue. There was a discussion on the venue for the 2017 meeting. There was a slight preference for Munich over Tokyo with August suggested as the most likely date. 11.2 Any other business None was raised. 12. Adoption of Resolutions (N2021) Resolutions LV1, LV2, LV9, LV10 and LV11 were approved by unanimous acclaim. Resolution LV3 passed after a vote - 12 yes, 2 no, 1 abstain. Resolutions LV4 to LV8 were approved by unanimous consent. There was a discussion on the future of the WG5 document archive, the website and the email systems but there was insufficient information available to make useful proposals. It was thought that the position would be clearer following the SC22 meeting in September. 13. Adjournment The meeting adjourned at 11:05 on Friday, June 27, 2014.