ISO/IEC JTC1/SC22/WG5 N2132 Minutes of Meeting of ISO/IEC JTC1/SC22/WG5 Hosted by DIN, the German Member Body for JTC1/SC22 in Garching, Bavaria, Germany June 26-30, 2017 List of Participants: John Reid (JKR Associates, UK) convenor Dan Nagle (NCAR, USA) PL22.3 chair Reinhold Bader (Leibniz Supercomputing Centre, Germany and DIN) Daniel Chen (IBM, Canada) Tom Clune (NASA, USA) Malcolm Cohen (NAG, UK) Brian Friesen (Lawrence Berkeley National Laboratory, USA) Gary Klimowicz (NVIDIA, USA) Tom Knox (Kernelyze, USA) Steve Lionel (Kernelyze, USA) Bill Long (Cray, USA) Toon Moene (Gnu Fortran, Netherlands) David Muxworthy (British Standards Institution, UK) Anton Shterenlikht (University of Bristol, UK) Nathan Sircombe (AWE, UK) Van Snyder (Caltech/JPL, USA) Jon Steidel (Intel, 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/docs.html 3. INCITS/PL22.3 is abbreviated throughout to PL22.3 and its papers are referenced as J3/17-nnn. They are available from http://www.j3-fortran.org/ 1. Opening of the Meeting The meeting opened at 09:00 on Monday, June 26th 2017. 2. Opening business 2.1 Introductory remarks from the Convenor The convenor welcomed participants, especially those who were attending WG5 for the first time. He noted that this was the eighteenth, and last, occasion in which he and Dan Nagle would jointly chair a Fortran standards meeting. 2.2 Welcome from the Host Reinhold Bader welcomed participants on behalf of DIN and the Leibniz Supercomputing Centre (LRZ). 2.3 Local arrangements Reinhold Bader described local arrangements and invited participants to dinner at the Gasthof Neuwirt in Garching on the Wednesday evening. There would be a presentation on the LRZ visualization infrastructure and a guided tour of the LRZ computational facilities after the close of the meeting on Tuesday and Thursday respectively. 2.4 Appointments for this meeting The drafting committee would be Reinhold Bader, Daniel Chen, Toon Moene, David Muxworthy (chair), Jon Steidel and Masayuki Takata. David Muxworthy would act as secretary and John Reid as librarian and as editor of the Disposition of Comments document in response to the CD ballot. 2.5 Adoption of the agenda [N2116] The agenda was adopted. 3. Matters arising from the minutes of the Boulder 2016 Meeting [N2109] There were no items not otherwise on the agenda. 4. Status of Boulder 2016 Resolutions [N2108] With regard to resolution B4 and the Strategic Plan for WG5 (N2016) the convenor said it would be necessary during the meeting to decide whether or not to produce a second CD. In either case the ultimate publication date of July 2018 would be met. 5. Reports 5.1 SC22 Matters (Convenor) John Reid reminded members that he was resigning as WG5 convenor at the SC22 plenary meeting in August 2017. There would also be a change in the SC22 convenorship with David Keaton, from the USA, taking over at that meeting. A new requirement by JTC1, that all participants in working groups formally register with ISO (v. SC22 N5201) in order to attend a meeting, was considered inappropriate for this group. The culture of the group was to be as inclusive as possible and was not to exclude those who might contribute to the work. This matter would be taken up at the SC22 plenary meeting. 5.2 National Activity Reports Canada: The most recent SCC (Standard Council of Canada) quarterly meeting was in early June. The committee draft of the next Fortran standard was approved. Germany: Reinhold Bader was now formally a DIN committee member and attended one DIN meeting per year. He was the only active member for Fortran in Germany, other members having ceased Fortran activity. Japan: The Fortran working group had eight members and met face to face five or six times a year. The translation work for F2008 had not proceeded far enough for publication by the end of its life. Japan voted for the CD ballot with some editorial comments. A WG5 meeting in Tokyo was planned for 2019. Netherlands: The Programming Languages Committee of NEN had ceased to function; it was not clear whether it had been formally disbanded. Toon Moene continued as an individual expert member of WG5. UK: A report was in N2128. USA: The main activity was that of PL22.3. 5.3 Report from Primary Development Body (INCITS/PL22.3 Chair) PL22.3 was charged with developing the draft standard and had done so. 5.4 Reports from other Development Bodies (Editors/Heads) There were currently no contributing development bodies. 5.5 Liaison Reports: NCITS/J11 (C): Dan Nagle C had not been followed in detail since the Cplex project had terminated. MPI: Bill Long The fault tolerance proposal was still in limbo. Enough cases were discussed during the December meeting to challenge the ULFM proposal and the working group was instructed to go back and work on them again. The next major item for consideration was MPI Sessions. UPC: Damian Rouson No report. IFIP/WG2.5: Van Snyder A meeting had been proposed for this summer but would not take place. OpenMP: Bill Long There was a week-long face to face meeting in the previous month, where progress was made on various topics. OpenMP 5.0 was still on track to be released at SC'18. For this year's SC, a "beta" preview draft of OpenMP 5.0 -- TR6 would be released. The two major new features that were expected for TR6 were memory directives and accompanying runtime APIs, and OMPD (an interface for third-party debuggers and the OpenMP implementation). There were also some additional tasking features that were expected to be introduced, some clarifications on the use of Fortran allocatables/pointers with devices, and probably some language to account for "newer" C++ features (e.g. lambdas). The memory directives would include the directives described in the released TR5 document, and the API would probably also define some predefined allocators for memories optimized for various attributes (e.g. bandwidth-optimized) WG9 (Ada): Van Snyder WG9 and Ada Europe had a joint meeting in conjunction with a conference on software reliability, in Vienna, during the week of 16 June. Van Snyder was requested to ask Randy Brukardt or Erhard Plodereder if WG9 had prepared separate WG23 annexes for Ada as a whole, and SPARK, the high-reliability subset of Ada. WG23 (Vulnerabilities): Dan Nagle There was a WG23 meeting the previous week where work continued on the base document ISO/IEC TR24772-1. Work would resume on the Fortran annex (ISO/IEC TR24772-8) when time was available. Members of WG5 were asked to volunteer to help the progress the Fortran part of TR24772. Cplex had completed its contribution and their document had gone to WG23 for inclusion in the C part. OpenACC: Gary Klimowicz OpenACC 2.6 is hoping to complete sometime in 2017. Proposed features include: - Manual Deep Copy: a relatively minor change to the data clause behavior that would allow a user to perform a manual deep copy. OpenACC is continuing to investigate true deep copy, but are looking for a prototype implementation before proposing adoption. - A serial compute construct for use between parallel compute constructs that helps avoid unnecessary data movement. - Query routines that allow the application to detect available devices and properties of the devices. - Support for an error callback routine so an application can gracefully exit if there is a fatal error in device code. - An if and if_present clause on the host_data construct to direct calls to either host addresses or device addresses, depending on where the data resides. - A no_data clause on compute constructs that would behave like present-or-don’t-create. - Data clause modifiers, such as copyin(readonly:x) that should allow for more aggressive code optimization. OpenACC anticipates a minor release 8-12 months following 2.6 where they hope to address deferred topics: - Partially shared memory - Reductions across multiple loops - Asynchronous data constructs and the “present” table - Reductions as a data clause - Several others More information is available at http://www.openacc-standard.org/. The current version of the proposed specification can be found at OpenACC 2.6 Specification Update (June 2017) (https://www.openacc.org/sites/default/files/inline-files/ OpenACC_2pt6_Specification_Update_ISC17.pdf). Following agenda item 5 discussion alternated between items 6, 7, 9 and 10, principally in PL22.3 mode, for much of the remainder of the meeting. This allowed for review of each day's progress on the various topics overnight. Final versions of documents were agreed on the last day of the meeting. 6. Consideration of the comments from the ballot of the CD revision of the Fortran standard and construction of responses. 7. Construction of edits to the CD revision of the Fortran standard to correspond to the CD ballot responses and any other changes deemed desirable. 8. Consider the Fortran defect reports (interpretations) in J3/17-006. 9. WG5 Business and Strategic Plans 9.1 Goals for 2017-2020 10. Start planning for the further revision of the Fortran standard. Agenda item 8 was not pursued at this meeting. There follows an outline record in chronological sequence of discussions in WG5 mode. [PL22.3 plenary and subgroup sessions from 09:55 to 16:30] Arrangements for the evening were announced. [PL22.3 plenary session from 16:32 to 16:45] The meeting adjourned at 16:45 ........................................................................... Tuesday The meeting resumed at 09:00. [PL22.3 plenary and subgroup sessions from 09:00 to 17:30] During the plenary session Dan Nagle gave a short presentation on development work at NCAR on the h2m Autofortran Tool - a tool to translate C headers to Fortran modules to let Fortran programs use C library functions (https://www2.cisl.ucar.edu/sites/default/files/ Garnet_Liu_Presentation.pptx). The convenor reported that Damian Rouson, the host for the 2018 WG5 meeting, proposed to contact the meeting via Skype on Wednesday morning at 09:00. Also members were invited to review draft versions of the updated WG5 Strategic Plan and a preliminary Disposition of Comments document which were on the meeting website. [PL22.3 plenary session from 17:32 to 17:45] The meeting adjourned at 17:45 and was followed by a presentation by Rainer Oesmann on the LRZ Centre for Virtual Reality and Visualisation (V2C). ........................................................................... Wednesday The meeting resumed at 09:00. There was a discussion with Damian Rouson on detailed arrangements for the 2018 WG5 meeting. Richard Gerber had been co-opted as co-host; the meeting would be held at Lawrence Berkeley Laboratory; it would be possible to stay in the LBNL Guest House but it was too early to reserve this accommodation. A file giving further information was available on the meeting website and was subsequently allotted number N2136. The reunion meeting for former Fortran committee members was being organized separately. [PL22.3 plenary session from 09:30 to 10:30] During the session there was a WG5 straw vote on PL22.3 paper 17-155r1. Should a PROTECTED TARGET be prohibited from being an initial-data-target? Vote: yes 16 - no 0 - abstain 2. Should the prohibition as a data-target in a structure constructor be a constraint rather than a requirement? Vote: yes 18 – no 0 – abstain 0. The vote affected various alternatives proposed in the paper. After reconvening in WG5 mode there was a discussion on the first draft of the revised Strategic Plan for WG5. The essential choice was whether or not to produce a second CD. In either case final publication of the standard was expected to be earlier than the present forecast date of July 2018. It was decided that the window for further technical changes had closed and that the draft standard at the end of the meeting should be the DIS, not a second CD. There was a review of the first draft of a Disposition of Comments document to respond to the CD ballot. Minor edits were suggested. [PL22.3 subgroup and plenary sessions from 11:30 to 15:10] Members were invited to review updated drafts of the WG5 Strategic Plan and Disposition of Comments document for discussion the following morning. It was suggested that the outline schedule for the reunion meeting in June 2018 could be: Saturday evening: dinner Sunday morning: presentation(s) Sunday afternoon: expedition, possibly to Alcatraz Sunday evening: dinner There was a review of the latest draft of the resolutions. Minor edits were proposed. Under agenda item 10 there was an hour-long discussion on the criteria to be used in planning the next revision of the Fortran standard. This was in the context of the revised Strategic Plan being developed at the meeting and the discussion paper N2126 (A Strategy for Reckoning the Content of the Next Revision). The debate was chaired by Steve Lionel as in-coming WG5 convenor. There was general concern that the standard had moved ahead of the language in compilers available to users, giving the impression that Fortran was slowing down. Moreover different compilers implemented different subsets of the standard, thus inhibiting both development and portability of programs. There was a need to investigate what Fortran users wanted to do in the language which was not currently possible and to find out what features in other languages were actually used, not simply talked about. Fortran was said to be the only language which built interoperability into its standard. New features should be adopted only if they had long-term viability and features which might have only a minor benefit for users but a negative effect on compiler construction or performance should be avoided. The debate was adjourned, to be continued the following day. Steve Lionel thanked John Reid for helping with ISO procedures with regard to the change of convenorship. The Project Editor advised that the latest version of the draft standard, which incorporated edits from papers approved earlier in the day, was accessible and asked that any errors be brought to his attention. The meeting adjourned at 16:35 and was followed by the meeting social event at the Gasthof Neuwirt in Garching. Entertainment there was provided by "Classic goes Bluegrass" with the host as lead fiddle. ........................................................................... Thursday The meeting resumed at 09:00. Updated drafts of the resolutions and the Strategic Plan were examined and minor changes suggested. Members were asked to check the latest version of the Disposition of Comments document and to report any errors to the convenor. The convenor's report to SC22 was discussed and minor edits were suggested. [PL22.3 plenary session from 09:20 to 09:55] A subgroup of three, Bill Long, Dan Nagle and John Reid, was appointed to check that the edits arising from this meeting were incorporated correctly into the proposed DIS (N2134). The Project Editor advised that hyperlinks in the document would be black, that is indistinguishable from normal text, and that the DIS would not have line numbers. The discussion on planning the next revision of the Fortran standard resumed, again chaired by Steve Lionel. Scheduling was debated, resulting in edits to the Strategic Plan. It was agreed that a web survey of users would be instituted, organized by Steve Lionel. The suggestions would first be processed by PL22.3 at its October meeting. The survey would be closed prior to the February 2018 PL22.3 meeting and further processing done at that time. At the June 2018 WG5 meeting a first cut at a work plan would be produced. Final versions of the Strategic Plan and Disposition of Comments documents were uploaded to the meeting website. 11. Closing Business 11.1 Future meetings Future meetings had been discussed earlier in the meeting. 11.2 Any other business None was raised. 12. Adoption of Resolutions [N2131] Resolutions G1 to G5 and G10 to G12 were approved by unanimous acclaim and resolutions G6 to G9 were approved by unanimous consent. Responding to resolution G1 John Reid, the retiring convenor, said he had much enjoyed working with WG5 members over the years, thanked them for the support they had given him and said that he proposed to continue working with Fortran as an ordinary WG5 member. 13. Adjournment The WG5 meeting adjourned at 12:15 on Thursday, June 29, 2017 and was followed by tours of the LRZ computing facilities conducted by Volker Weinberg and Reinhold Bader. The PL22.3 meeting had adjourned at 09:55.