BRITISH COMPUTER SOCIETY - FORTRAN SPECIALIST GROUP
Minutes of AGM held at BCS Headquarters on Monday 6 April 1981
Present: G Harding (Chairman) ECMWF
P A Clarke Rothamsted Experimental Station
J L Dyke Huntingdon Research Centre
J Roberts-Jones Liverpool City Council
B Meek Queen Elizabeth College
D Muxworthy Edinburgh University
M Nunn C.C.T.A
G A Ruscoe D L B S S Ltd
D M Vallance Salford University
T L Van Raalte MOD
A Wilson ICL
J D Wilson (Secretary) Leicester University
1. Apologies for absence
None.
2. Minutes of previous meeting [2 February 1981]
Page 2, item (iii) amended to omit:
"(i.e. PARAMETER, COMMON, DIMENSION be combined)".
The minutes were then accepted.
3. Matters arising
All covered by the agenda for this meeting.
4. Election of Officers
Gary Harding resigned as Chairman. The Group thanked him for
his work during his period of office.
John Wilson was proposed for Chairman by Gary Harding, seconded
by P A Clarke. There were no other proposals and J D Wilson
was elected by a unanimous vote.
John Roberts-Jones agreed to stand as Treasurer again. He
was proposed by D Muxworthy, seconded by B Meek and elected
unanimously.
David Vallance was proposed as Secretary by J D Wilson,
seconded by G A Ruscoe and elected unanimously.
T L Van Raalte was proposed as Vice-Chairman by B Meek,
seconded by D Vallance and elected unanimously.
Gary Harding resigned from the Steering Committee owing to other
commitments.
At this point the new chairman and secretary assumed their duties.
5. Treasurer's Report
The treasurer made an informal progress report as the formal
report was not due until the next meeting.
The general costs of mailing and BCS services were discussed
in relation to the level of financial support from the BCS.
(There are approximately 130 people on the mailing list.)
It was agreed that the £2 charge for group membership for
non-BCS members should not be increased.
6. FORTRAN Forum
The date for the Forum is Monday October 12, 1981. Edinburgh
University are holding a similar event on Wednesday 14 October 1981
for the benefit of those from the north of England and Scotland
who would find in difficult to get to London. The two events
will attempt to share publicity.
Brian Meek was thanked for his success in booking a room
provisionally at the University of London Institute of Education.
This booking will be formally confirmed after July when the
Institute agrees charges for the next session.
The Elvin Hall (seating at least 150) has been booked for the
meeting (current charge £85) and the Jeffrey Hall (current charge £55)
for lunch. These charges are expected to increase by about 10%
after July. The Elvin Hall has very good audio-visual facilities, Action:
should these be required. J D Wilson
Third party liability had not been discussed. The Chairman
will ask the BCS to write to the Institute of Education and
confirm that we are covered by the BCS's insurance.
It was agreed that any institutional member of the BCS could
send any number of delegates to the forum at the rate for BCS
members.
An attendance of 100 was estimated - the actual cost per
delegate will not be available until the booking is confirmed
in July.
D Muxworthy stated that at least 4 X3J3 members are interested
in attending from the USA. There might ultimately be as many
as 10.
The morning of the Forum will be devoted to speakers who are
Fortran 77 users; the afternoon will allow X3J3 to discuss
their current work.
Mike Metcalf (CERN), Rex Gibson (Weather Centre) and John Action:
Larmouth (Salford University) were to be approached as possible J D Wilson
speakers. G Harding
D M Vallance
The possibility of approaching Computer Weekly for support
was discussed.
A working group consisting of the Chairman, D Muxworthy (who
will liaise with X3J3), Alan Wilson, Gary Harding, T L Van Raalte
and John Roberts-Jones was set up to organise the Forum.
D Muxworthy will draw up a draft programme. Action:
D Muxworthy
J Roberts-Jones will estimate the cost of publicity based on Action:
the previous Forum. J Roberts-Jones
The Chairman will investigate the catering charges. Action:
J D Wilson
Initial publicity will be distributed as soon as possible.
Alan Sutcliffe of the BCS will be informed of the proposed Action:
forum at the meeting of the Specialist Groups Committee on J D Wilson
14 May 1981.
7. Future meetings of the Group
The following dates were agreed for future meetings:
12 October 1981 FORTRAN FORUM
30 November 1981 Possibly at University of Salford
5 April 1982 AGM
The Chairman will contact John Murchland as a possible speaker Action:
for 1 June. The secretary will contact David Law (NCC) as a J D Wilson
possible speaker for September. Alan Wilson will approach D M Vallance
Professor Parkinson at the QMC DAP Centre regarding a possible A Wilson
meeting there at which Professor Parkinson would speak in the
afternoon.
8. BCS Business
Members have been asked for recommendations for distinguished
or honorary fellows of the BCS. Such recommendations should
be sent to the Chairman or Secretary of this Group.
Gary Harding reported from the recent meeting of the Specialist
Groups Board. Two new groups have been affiliated: Computing
in Psychology and the Eurographics Association. Changes have
been proposed in the byelaws regarding specialist groups which
will be discussed at the next meeting of the Board.
9. Report from March meeting of X3J3 at Austin, Texas
Alan Wilson reported on the last X3J3 meeting but noted as
he himself was heavily involved in array processing discussions
his notes were unfortunately brief.
The macro facility had been discussed again as had precision
proposals. The question of how subprograms could be compiled
in a variety of forms was still unanswered.
The concept of independent compilation is still in conflict
with GLOBAL.
Walt Brainard has produced a composite document X3J3/S6 'Proposals
approved for Fortran 8X' which will be updated as new proposals
are approved.
10. ISO Programming Languages Meeting
Brian Meek reported that the BCS standards committee had put
forward a request for substantial sponsorship for the meeting.
The Chairman was asked to write to the BCS Secretariat
recommending support.
A £50 contribution from the Fortran Specialist Group funds
was proposed by A Clarke, seconded by G A Ruscoe and agreed Action:
unanimously. J Roberts-Jones
11. Any Other Business
Salford University are holding a 1-day seminar entitled
'Experiences with Fortran 77' on Monday 8 June 1981. The
fee (£16.50) includes coffee, lunch and a copy of a newly
published Fortran 77 textbook. The speakers include Alan Wilson
and John Larmouth. Full details are obtainable from
Mrs Roberts, Short Courses, Registrar's Department, University of
Salford, Salford M5 4WT.
Mr Wigg has produced a program to analyse the structure of a
Fortran program and is anxious to find a Fortran user, preferably
in the South East, who might be interested in assessing its
usefulness. Those interested are invited to contact him at
72 Brattle wood
Sevenoaks
Kent
TN13 1QU Tel: (evening) 0723 - 55446
Afternoon Session
FORTRAN 8X ARRAY PROCESSING PROPOSALS - ALAN WILSON, ICL
Appendix A of these minutes is a summary based on an article
written by Mike Metcalf in CERN Newsletter 155.
APPENDIX A FORTRAN 8X ARRAY PROCESSING PROPOSALS
There is a clear need in mathematical programming language to provide
a simple means of manipulating vectors and arrays. This need will grow
with the increasing use of vector processors. The discussions on which
facilities to provide are far from complete, but seem to be working towards
one set of basic features to be incorporated into the core, and a further
set to form part of a language extension.
Basic array processing
The operations and intrinsic functions are extended to apply to whole
arrays on an element-by-element basis, to produce a result of the same
shape (dimensionality and dimension size) as the operand(s). For
operations involving two operands, the operand arrays must be conformable
(of the same dimensions), but a scalar is defined as being conformable
to all arrays, and is 'broadcast' as necessary. A combined example is
thus.
DIMENSION A(20,10),B(20,10),C(20,10)
.
c=3.*A*SQRT(B)
The WHERE statement and WHERE block allow testing on an element-by-element
basis:
WHERE (B.NE.0) C=A/B
OTHERWISE
C=A
B=EPSILON(X) ! EPSILON is an environmental enquiry function
END WHERE
In order to permit the addressing of array sections a new, as yet
unspecified symbol, is introduced (here '*' is used). It permits, for
example, the columns and rows of a two-dimensional array to be selected
in the following manner, where the minus sign indicates that the elements
are to be taken in reverse order:
using DIMENSION A(-4:0,7) ! FORTRAN 77 array notation
then A(-3,*) selects the second row of A
A(*,3) selects the third column of A
A(-1,-*) selects the fourth row of A, backwards.
A triplet notation similar to DO-loop parameters will permit references
to non-contiguous elements of an array:
A(0:-4:2,1:7:2) selects in reverse order every second element
of every second column.
Extended array processing
This topic would require a great deal of space to develop, and this
summary is necessarily brief. The following lists of proposed functions
should just serve to give a feeling for the present ideas.
a) Intrinsic functions return scalar values from arithmetic arrays:
SUMN, PROD, MAXVAL, MINVAL, DETERMINANT, DOTPROD
and from logical arrays
COUNT, ANY, ALL.
b) Informative functions return the
RANK, EXTENT, SIZE, LBOUND, UBOUND
of an array.
C) Generative operators fill a vector with a
SEQuence of numbers, defined by 3 DO-loop like parameters
ALTernate sequences of .TRUE. and .FALSE. values, where
the number of sequences and lengths are
specified by 3 parameters.
In the case of array it is possible to
SPREAD values into an array of higher dimensionality
REPLICATE them within an array
MERGE 2 arrays according to the logical values of
third array
DIAGONAL i.e. produce a unit matrix.
d) Shift operators can perform
EOSHIFT (end-off) and
CSHIFT (circular) shifts.
e) Finally, matrix operations such as MATMUL, INVERSE and TRANSPOSE
are proposed.