BRITISH COMPUTER SOCIETY - FORTRAN SPECIALIST GROUP


   Minutes of meeting held at QMC Mile End Road on Monday 7 September 1981


        Present:    J D Wilson (Chairman)      Leicester University

P A Clarke                 Rothamsted Experimental Station

M R Dolbear                BP International ltd

J L Dyke                   HRC

E Gorczynski               BP Research Centre

B Meek                     QEC

K Normington               Coventry Polytechnic

M Nunn                     CCTA

T L van Raalte             MOD

P J Wallis                 Bath University


1.      APOLOGIES FOR ABSENCE


        Apologies were received from D Vallance, D Muxworthy and G Harding.


2.      MINUTES OF PREVIOUS MEETING [1 June 1981]


        Item 5.1.4 Appendix B was not the Fortran Specialist Group's annual

        report


        Item 7.2 For 'accord' read 'are wary of'.


3.      MATTERS ARISING


3.1     ISO programming languages meeting


        Brian Meek said that as a sponsoring body, a representative from the

Fortran Specialist Group would be welcome at the Monday meeting.


        The Treasurer is still unavailable to complete the transfer of the

agreed donation.


                                                Continuing Action - J Roberts-Jones


        Brian Meek acknowledge the favourable response of the BCS secretariat

to the letter of support written by the Fortran Specialist Group.


3.2     The Fortran Specialist Group Annual Report


        John Wilson said in answer to a question by Mike Dolbear that the annual

report would accompany the current minutes. (See Appendix A of these minutes).


4.      BCS BUSINESS


4.1     Specialist Groups Board


        There will be a meeting of the Specialist Groups Board on 10 September

at 2 pm in BCS HQ.


4.2     Publicity for Fortran Specialist Group meetings


        The absence from Computing a notice of the current meeting was mentioned.

In answer it was said that notice of the year's meetings is being sent to BCS

to ensure better circulation in future.


5.      FORTRAN FORUM


      The appropriate BCS board has not yet met to give official approval to

the event. Replies made by John Wilson to BCS enquiries in June have not

elicited any unfavourable response.


        Official confirmations from the Institute of Education for the booking

of the hall and for catering have been received. (The Jeffry Hall and Foyer

have been booked all day and Committee Rooms l and 2 for lunch, which will be

a buffet with wine. There will be an ohd projector).


        Deposits are now needed. In the absence of the Treasurer, the Chairman

will if necessary, make alternative arrangements for payment and reclaim the

amount later.                                                    Action - J Wilson


        David Muxworthy has made minor changes to the draft agenda dated

l9 August 1981. Mr Metcalf will not be available to give a user's viewpoint.

Brian Meek suggested Manchester as an alternative source of speaker, and
offered to contact Richard Field at URCC immediately.

                                                                 Action - B Meek


        John Wilson will approach Mike Ellis at Sheffield.


                                                                 Action - J Wilson


        It was agreed that the slot could be absorbed by ANSI speakers if necessary.


        Alan Clarke suggested organising a literature stand displaying material

on Fortran standards and video-courses. Brian Meek recommended asking Dillon's

University Bookshop to supply this. John Wilson will check what effect this may

have on the arrangements with the Institute of Education.

                                                                 Action - J Wilson

        Alan Clark will approach Dillons.                        Action - P A Clark


        The BCS will give a quarter page advert (and application form) in the

September 17 issue of Computing. Notifications have already gone to the Fortran

Specialist Group mailing list. John Wilson has contacted all British

Universities.  More publicity is urgently needed and all concerned with Fortran

are urged to give the Forum as much publicity as possible.


6.      X3J3 BUSINESS


6.1     X3J3 Meeting


        The last X3J3 was reported as largely concerned with clarification.

The Obsolete Features and Array Processing Modules are to be called the

Transition and Scientific modules respectively. A likely form for Fortran 8X is

the core plus these two modules.


6.2     Documentation


        Jean Adams expects the Draft definition to be written next summer. In

reply to a question on other useful documents on F8X John Wilson said that

ANSI's intentions are not yet known. They may wish only the Draft to be given

publicity.


        Loren Meissner has produced a useful account of his expectations for F8X

in Weather Vane. Standing Document Sl has also been produced.


6.3     Conformity


        There is nothing further to report. Its present status in X3J3 is not known.


7.      ANY OTHER BUSINESS


7.1     Peter Wallis, chairman of the BCS Software Engineering Group is working

on an Ada-to-Fortran conversion and would like to hear from anyone interested

in helping with the Fortran end. His address is Dr P J Wallis, School of Maths,

University of Bath, Claverton Downs, Bath, BA2 7AY


7.2     Mike Metcalf of CERN has written apologising for his absence from the

Fortran Forum, and asking for information on ICL, Univac and Burroughs

compilers for F77.  Mike Dolbear may be able to help with Univac and

K Normington with Burroughs.


7.3     There is a Vector Processing seminar on 28 September to 2 October in

Bologna, Italy.


8.      FUTURE MEETINGS


12 October      Fortran Forum London

14 October      Fortran Forum Edinburgh

30 November     U.of Salford. Brian Shearing is to talk on

                Portable Fortran

1 February      BCS HQ

5 Apr           BCS HQ


9.      AFTERNOON SESSION


        Professor Parkinson of QMC gave a talk entitled 'Parallel Processing -

what is it.'. Starting from the text that 'all computers are parallel but some

are more parallel than others' he developed a taxonomy of computer types :

serial, vector processors and processor arrays. Using this he made, inter alia,

the points:-


1. Vector processors and processor arrays need different algorithms from

   serial machines and from each other.

2. The choice of algorithm depends on the depth of parallelism provided

   by the computer

3. Because programs represent algorithms rather than 'what one wants to

   do efficient portability does not exist.

4. Data structures are crucial if the potential of parallel processing

   is to be exploited.

5. The current extensive use of assembly language programs on parallel

   machines suggests that users find existing compilers to be

   unacceptably inefficient.

6. One only gets 'value for money' from investment in parallel machines

   if one is prepared to reprogram problems appropriately. If not then

   just buy more serial power.

7. The floating-point speed is becoming less clearly a measure of machine

   speed. Access to the pipeline: for example, can seriously degrade

   performance.

8. There is need of much higher peripheral rates to match the power of

   processing.




APPENDIX A



    British Computer Society - FORTRAN Specialist Group


Annual Report for 1980/81



        The Group met 5 times during the year, each time at BCS Headquarters,

with an average attendance of 17. The meetings followed the traditional

formula of a business/reporting session in the morning followed by a talk by

an invited speaker in the afternoon. The following is a list of speakers

and topics.


Dr. B. Smith, Argonne National Labs., USA and ANSI X3J3 member

        - Precision Proposals for FORTRAN 8X.


Dr. J. Larmouth, Salford University

        - Portability of FORTRAN 77.


Dr. L. Meissner, Lawrence Berkeley Labs., USA and secretary of ANSI X3J3

        - US Dept. of Energy FORTRAN language requirements.


S. J. Hague, NAG Central Office, Oxford

        - Tools for Numerical Software Engineering.


Dr. A. Wilson, ICL and ANSI X3J3 ember

        - Array Processing Proposals for FORTRAN 8X.


        Contact has been maintained with the XJJ3 (Fortran) Committee Of ANSI.

Two members of the Group (B. Ford and D. Muxworthy) attended X3J3 meetings in

the States (at their own expense) and a proposal for a "conformity" or

"completeness" module was put to the Committee. This was favourably received

and is being developed.


        A number of members of the group attend a Fortran Experts Meeting

organised by the ISO in Amsterdam in October 1980. Presentations were made

by A. Clarke, D. Muxworthy, J. Reid and L. Schoenfelder.


        Plans are currently in hand for a FORTRAN FORUM to be held in London

on October 12th 1981 and in Edinburgh on October l6th l98l. It is hoped to

have presentations by a number of ANSI members who will report on the status

of the next Fortran standard.






J. D. Wilson

Chairman, BCS Fortran Specialist Group