ISO/IEC JTC/SC22/WG5-N1691 Minutes of Meeting of ISO/IEC JTC1/SC22/WG5 Hosted by the British Standards Institution August 6-10, 2007 List of Attendees: WG5: John Reid (Convenor) Dan Nagle (J3 chair) Ian Chivers (UK Fortran Group, UK) Malcolm Cohen (NAG, Oxford, UK) Aleksandar Donev (J3, LLNL, USA) Dick Hendrickson (J3, USA) Michael Ingrassia (J3, Sun Microsystems, USA) Bill Long (J3, Cray, USA) Jeanne Martin (J3, USA) Steve Morgan (University of Liverpool, UK) Secretary Toon Moene (J3, Gnu Fortran, Netherlands) David Muxworthy (UK Fortran Group, UK) Shivarama Rao (Hewlett Packard) Lawrie Schonfelder (UK Fortran Group, UK) Jane Sleightholme (UK Fortran Group, UK) Van Snyder (J3, Caltech/JPL, USA) Masayuki Takata (Edogawa University, Japan) Jim Xia (J3/IBM, Canada) Guests: Greg Colyer, Culham Laboratories (Monday only) Apologies for absence were received from: Craig Rasmussen (J3, USA) Matthijs van Waveren (Netherlands) 1. Opening of the Meeting The meeting opened at 0900. 2. Opening Business 2.1 Introductory Remarks from the Convenor John Reid welcomed everyone to the meeting. He congratulated J3 on the magnificent job they had done in producing the draft standard. He observed that the UK had expressed serious reservations about the current schedule and the contents of the document and felt that it would be useful to explore these issues during the day. 2.2 Welcome from the host Jane Sleighthome and Ian Chivers welcomed everyone to London and BCS Headquarters. 2.3 Local arrangements. A reception sponsored by the BCS (arranged by Peter Crouch) and a trip on the river for Wednesday afternoon had been organized. 2.4 Appointments for this meeting David Muxworthy would act as chairman of the drafting committee. Steve Morgan would act as secretary. 2.5 Adoption of the agenda (N1672) The agenda was adopted. 2.6 Approval of the minutes of the Fairfax meeting (N1654) The minutes were approved with no objections. Malcolm Cohen said that he would prefer meetings to start on Sunday and end on Friday morning. 3. Matters arising from the minutes There were no matters arising not covered by subsequent agenda items. 4. Status of the Fairfax Resolutions (N1654) The second corrigendum had been completed. 5. Reports 5.1 SC22 Matters John Reid reported that he had raised the problems WG5 experienced with processing documents, that the chairman of SC22 had visited Geneva to discuss these with Keith Brannon. Keith has agreed to our suggested procedure and will attend the September meeting of SC22 to pursue the matter further. 5.2 National Activity Reports (Heads of delegations) N1682 UK National Activity Report (Muxworthy) N1684 US National Activity Report (Snyder) N1685 Japan National Activity Report (Takata) Jim Xia noted that Canada agreed with a proposed 2 year slip in publishing the standard and preferred that co-arrays be included. 5.3 Report from Primary development body (INCITS/J3 Chair) Dan Nagle noted that the draft standard (N1678) had been produced. 5.4 Reports from Other Development Bodies There were none. 5.5 Liaison Reports INCITS/H2(SQL) - not much to report INCITS/J11(C) - Craig Rasmussen was not present to report. Dan Nagle noted that the C committee encouraged compatibility of Fortran with the C special functions standard that was being developed. John Reid thought this was a low priority issue for this meeting. IEEE 754 revision - no comment IFIP/WG5 - nothing to report. OpenMP - no report (Matthijs van Waveren not present) OWG-V - Dan Nagle noted that he was acting as liaison to OWG-V (Working Group on Vulnerabilities). This initiative comes from requirements by US government (with some support from the UK) that software standards should deal with security in relation to such issues as cyber terrorism etc. (a web site reference is in N1686). Other language standards groups such as C, C# and Ada attend. 6. Review the current working draft (N1678) for the revision of Fortran 2003 John Reid asked David Muxworthy to present the UK's proposals (N1682) for delaying the standard and for reducing its size (removing Co-arrays, Macros, Bits). There followed a wide-ranging discussion on the issues. Compiler vendors were asked for their initial opinions and then non-vendors were invited to comment. Straw vote taken: Do we want to change the status from that specified in the Fairfax resolutions? Yes 10 No 2 Undecided 5 Straw vote taken: Keep everything 2 Reduce the size of the revision 7 Delay the revision 1 Reduce and delay 7 Undecided 1 There followed a discussion on what form a TR on co-arrays would take. Monday p.m. There was a long discussion on the schedule and features. Return to co-arrays. Straw vote taken: Keep Co-arrays in next revision 13 Opposed 1 Undecided 3 Straw vote taken: There should be no significant delay in the standard 5 There should be some delay 5 Undecided 8 There followed a discussion on the Bits proposal. It was pointed out that the bits proposals had several components e.g. the introduction of the new type (which some members objected to on principle) and the addition of features and intrinsics (which had been specifically requested by users). It may be useful to keep some of these features. Straw vote taken: Keep everything as is (in draft) 0 Keep type but restrict length 3 Keep intrinsic/other features but remove bits type 10 Remove all the new bits features 0 Undecided 3 There followed a discussion on the length of delay for publishing the next standard favoured by members. Straw vote taken: For a one year delay 6 Against a one year delay 5 Undecided 6 There followed a discussion on the Macros feature. Straw vote: Keep in standard 1 Move to TR 5 Leave out completely 5 Undecided 6 Straw vote: Remove from next standard Yes 10 No 1 Undecided 6 There followed a discussion on the new block features. Straw Vote: Keep blocks in the standard Yes 13 No 2 Undecided 3 Tuesday 9:45 Bill Long presented the issues concerning which of the bit features to keep in the standard. He identified several features for removal illustrated by the following :- 1. Assignment or initialization X = Z"ffffff" 2. I1 = I2.XOR.I3 I1 = I2.AND.Z"FF" 3. WRITE(*,'(Z18.16)')R 4. I1 I2 or BLT(I1,I2) 5. IAND(I2,Z"FF") Straw votes were taken on whether each feature should be retained :- Vote on feature 1: Yes 8 No 9 Undecided 1 Vote on feature 2: Yes 6 No 6 Undecided 5 Vote on feature 3: Yes 11 No 2 Undecided 5 Vote on feature 4: Yes 16 No 0 Undecided 2 Vote on feature 5: Yes 8 No 2 Undecided 8 John Reid noted at this point that there was consensus that there should be some delay in the standard but that there was no consensus on the length of the delay. A straw vote was proposed: Should we have a 2 year delay with a TR for co-arrays? Yes 3 No 12 Undecided 4 Straw vote (4-way): Those for a delay of 6 months 8 Those for a delay of 12 months 8 Those for a delay of 24 months (no Co-Arr TR) 1 Undecided 1 Some discussion indicated that there was no great difference between the 6-month and 12-month options in practice. The meeting then returned to considering the Macros issue. The discussion indicated that a type 2 TR would be suitable. Straw Vote: Do we want to make Macros a TR (type 2) Yes 8 No 3 Undecided 6 At this point it was agreed that Aleksandar Donev would act as editor for the TR [but see later for outcome]. Wednesday 0900 John Reid presented a draft revised schedule (N1693-1) showing possible revised dates. The effect of all the changes would be a 10-month delay with final publication in June 2010. There was then a discussion on the UK's comments on co-arrays. There was some discussion about Note 2.10 in the draft standard. This note indicates that co-arrays are simpler to implement for homogeneous systems (although it does not preclude their applicability to heterogeneous systems). It was felt that the form of words could be improved. Straw vote: Make a change to Note 2.10 Yes 15 No 0 Undecided 3 There was then a discussion on reducing the co-array features. Straw vote: Remove co-arrays as structure components? Yes 0 No 8 Undecided 9 Remove the collective intrinsics? Yes 0 No 9 Undecided 8 Remove one or more collectives (decided by subgroup)? Yes 12 No 3 Undecided 3 Issue deferred to subgroup. Remove type(image_team) & TEAM=? Yes 6 No 6 Undecided 6 Convenor asked the subgroup to consider this further (see Thursday). Remove some of the many synchronization primitives? (choice by subgroup) Yes 8 No 4 Undecided 6 Issue deferred to subgroup. Malcolm Cohen (Editor) raised the issue of paragraph and line numbers in the draft document. Line numbers cannot be included in the final standard but paragraph numbers can be. Straw Vote: Should paragraph numbers be left in the final published standard? Yes 15 No 2 Undecided 1 Thursday 15:00 John Reid presented his latest draft of the schedule for publishing the standard (N1693-2) John asked if there were any comments - there were none. John asked if anyone opposed this schedule - there was no opposition. The schedule was therefore agreed and would become N1693. There was a brief discussion on the name of the standard - should it be F2008/9/10? Straw vote: Retain 2008 as name Yes 12 No 4 Undecided 0 There was a discussion on the radix feature. Straw vote: Keep the radix argument as it is? Yes 8 No 3 Undecided 6 TR on Macros Malcolm Cohen had some fresh thoughts on whether there should be a TR on Macros. A discussion ensued followed by Straw Vote: Make macros into type 2 TR? Yes 4 No 5 Undecided 7 This reverses the previous decision and so no TR on macros was initiated. Simplification of Co-arrays Bill Long presented the pros and cons of removing the team features of Co-arrays. Arguments for keeping the feature were:- 1. Removing the feature would cause complicated edits 2. There was a performance benefit 3. Program safety 4. Migration to future design 5. Image list may be huge Arguments for removing the feature :- 1. Make for a simpler standard 2. Easier to implement 3. Future design may be better 4. Fewer SYNC statements. 5. Not used often Straw Vote: Remove type(image_team), sync_team, form_team? Yes 0 No 8 Undecided 9 Bill Long presented options for reducing the number of image synchronization statements. Straw Vote: Keep SYNC_IMAGES and NOTIFY/QUERY? Yes 6 No 2 Undecided 8 Straw Vote: Keep both 6 Delete SYNC_IMAGE only 1 Delete NOTIFY/QUERY 1 Delete both 1 Undecided 8 Straw Vote: Based on the subgroup report, various options for reducing the number of collective intrinsic subroutines were discussed. Delete co_findloc? Yes 8 No 4 Undecided 4 Go further and also delete co_maxloc/co_minloc collectives Yes 3 No 5 Undecided 9 Go even further and also delete co_maxval/co_minval collectives? Yes 2 No 5 Undecided 10 Delete all but co_sum and co_product Yes 1 No 6 Undecided 11 Delete co_findloc? Yes 6 No 4 Undecided 6 The convenor declared that co_findloc should be removed. Friday 10th Aug 0900: There was a discussion on the fact that, although ENTRY had been voted as obsolescent at the Fairfax meeting, this had not been changed in the draft standard (i.e. the text had not been changed to small font). This was recognized to have been a human error in handling the paper trail. The convenor suggested that the action (to make ENTRY obsolescent) should be confirmed by the meeting. Straw Vote: Ask J3 to make ENTRY obsolescent in the draft standard? Yes 12 No 1 Undecided 0 Straw Vote: Co-arrays should be in the main standard? 10 Co-arrays should be a separate part or TR 5 Undecided 2 There was a discussion on the special mathematical functions being developed as a C/C++ standard (N1687, N1688) Straw Vote: Should WG5 consider some form of standardisation of ISO 31-11 (special functions)? Yes 4 No 5 Undecided 6 The Convenor agreed to explore working with the WGs for C/C++ regarding new standards (TRs) for special math functions whilst recognizing that the WG5 response was mixed. 7. Consider the Fortran defect reports (interpretations) in J3-006 See resolutions. 8. WG5 Business & Strategic Plans Bill Long is working on further interoperability TR. His formal approval as project editor should be sought. Report to SC22 is in N1692. There should be a request for further corrigenda. Japan will probably be able to make a decision on hosting the November 2008 meeting in September. The next WG5 meeting will be Feb 10-15 2008 in Las Vegas. 9. Closing Business The resolutions (N1690) were discussed and a few minor changes were suggested. 9.1 Future meetings Arrangements for the meeting in Tokyo - Nov 16-21, Tokyo, Japan. 9.2 Any other business There was some discussion of N1694 (response to N1674) Straw Vote: Accept N1694 as modified? Yes 15 No 0 Undecided 0 N1695 (revision of bits feature) was discussed and some minor suggestions were made. Straw Vote: Accept N1695 as amended? Yes 13 No 0 Undecided 2 The convenor noted that he would appreciate comments on N1692 (WG5 Business Plan and Convener's Report) 10. Adoption of Resolutions (N1690) L5 (in the draft document) was the only controversial one. The resolution stated: L5. Content of the revised Fortran standard That WG5 makes the following changes to the content of the revision: - type BITS is removed and ancillary bit manipulation facilities are revised as shown in WG5-N1695 - co-arrays are simplified by removal of CO_FINDLOC - intelligent macros are removed. The convenor ruled that this resolution should be voted on separately. Individual Vote: Accept L5? Yes 10 No 4 Abstain 1 Country Vote: Accept L5? Yes 2 (US, Netherlands) No 0 Abstain 3 (Japan, Canada, UK) Individual vote on the other resolutions: Resolutions L1-L4 & L6-L9 were passed with unanimous consent (16-0) Country vote on the other resolutions: Resolutions L1-L4 & L6-L9 were passed with unanimous consent (5-0). 11. Adjournment John Reid thanked the local hosts again for their excellent support during the week. The meeting adjourned at c. 1300.