BRITISH COMPUTER SOCIETY FORTRAN SPECIALIST GROUP


Minutes of a meeting held on

Friday, 8th February 1974 at

BCS Headquarters, 29 Portland

Place, London W1 at 10.30 am.


Present:    Mr I.D. Hill (Acting Chairman)    MRC Computer Unit

Mrs E. Aylmer-Kelly               University of York

Mr B.J. Banes                     Rolls Royce, Bristol Engine Div

Mr S.J. Bell                      Computer Weekly

Mr E.O. Bodger                    IBM, W.T. Systems Aid Centre

Mr P.D. Bond                      Philips Industries

Mr A. Bruce                       Government Communications HQ

Mr F. Chambers                    Logica Ltd

Mr P.A. Clarke                    Rothamsted Experimental Station

Mr J.C. Cullen                    B.P.

Mr M.R. Dolbeare                  B.P.

Mr J.P. Holland                   Tunnel Cement Ltd

Dr C.B.A. Price                   Honeywell I.S. Ltd

Mr K. St Pier                     G.E.C.

Mr T.L. van Raalte                A.W.R.E. Aldermaston

Mr A.J.H. Walter                  Atlas Computer Laboratory                

Mr D. Winstanley                  University of Birmingham

Mr D.T. Muxworthy (Secretary)     Edinburgh R.C.C.


Apologies   Mr B.H. Shearing (Chairman)       Alcock Shearing & Partners

for         Dr A.C. Day                       University College, London

Absence

ACTION

1.    APPOINTMENT OF           As Mr Shearing was indisposed Mr Hill agreed

ACTING CHAIRMAN          to chair the meeting.


2.    APPROVAL OF              The minutes of the meeting of 9th November 1973

MINUTES                  were approved. (The Chairman of the American

Standards Fortran Committee, X3J3, had asked for

a clarification of section 2: this should have

made clear that it referred to publication of

the eventual draft Standard, with the approval

of X3J3, and not to publication of the current

FORTREV document).


3.    MATTERS ARISING          a.    There was no information about any contact

with the editor of the Computer Journal

concerning publication of the draft          BHS

American Standard for Fortran.


b.    Mr Hill and Mr F. Parker of Newcastle

University had written to the editor of the

Journal Algorithms Supplement concerning

(errors in Algorithm 78 (August 1973 pp 273-

276); no reply had been received. Mr Hill

was to take up the matter with Mr Shepherd.  IDH


c.    Mr Clarke had sent his report on using the

U.S. Navy Fortran compiler testing programs

to Captain Hopper. She had replied that

responsibility for the programs had been

transferred to the Federal COBOL Compiler

Testing Service (Department of the Navy,

ADP Equipment Selection Office, Washington

DC 20376), and that the National Bureau of

Standards were understood to have a "much

better" set of Fortran tests. {The

contact is Mrs F.E. Holberton, National

Bureau of Standards, Bldg 225, Room BBQQ,

Gaithersburq, Md 20224).


4.    CHAIRMAN'S               This agenda item was omitted because of

REMARKS                  Mr Shearing's absence.


5.    PARTICIPATION            The Group had been invited to participate in

IN EUROCOMP              Eurocomp to be held at Brunei University on

May 13-17, 1974. It had been suggested that

the Group might make a presentation, as at

Datafair 73, or arrange a static exhibition.

It was decided that a presentation should not

be made unless the new draft Standard was

available, and little interest was shown in

mounting an exhibition, although expert help

was thought to be available from the organizers.


It was decided to appeal to members of the Group   All

to contact the Secretary if they wished to help

arrange a Fortran event at Eurocomp and to post-

pone discussion until the next meeting.


6.    SCHNECK-ANGEL            A member of the Group, Mr D.T. Hall, had

OPTIMIZING               suggested that the Group, per se, obtain a copy

COMPILER                 of the Schneck-Angel Fortran-to-Fortran Opti-

mizing Compiler (Computer Journal, November 1973,

pp 322-330). Mr Hall offered to help verify it

and suggested that the Journal Algorithms

Supplement might use it for published algorithms.


Mr Walter said that the Atlas Laboratory had

requested a copy of the program but had as yet

received no reply; they intended to verify it

and then to use it when appropriate. There was

no suggestion that it become standard. Subject

to agreement by the authors, the Atlas Laboratory

would make the program available to other

installations.


There was scepticism about the claims for improve-

ment in running times made for the compiler and it

was demonstrated that optimization at source level

could lead to an increase in the number of machine

instructions generated and executed. It was not

thought appropriate to bring the program to the

attention of the Algorithms Editor.


7.    BSI FORTRAN              There was some confusion over the status of

      ACTIVITIES               the Fortran representative on the BSI

Programming Languages Committee DPE/13.

Mr Shearing had been informed that he had

replaced Dr Day, other members of DPE/13

had been informed that Dr Day was no

longer a member but Dr Day himself had

received no communication. This was under-

stood to be related to Dr Day's

resignation from the BCS although by

membership of other societies with

reciprocal agreements there was no impedi~

ment to his representinq the PCS on

DPE/13.


It was decided that the Secretary should         DTM

seek a clarification from Mrs Leach.

There was some urgency in resolving the

matter as an ISO Fortran meeting was to be

held in May 1974. A vote was held to

record the meeting's opinion: if the Group

has the power to nominate or recommend a

member of DPE/13 (which was earlier under-

stood to be the case) then a motion that that

nominee be Dr Day was approved unanimously.


8.    ECMA FORTRAN             Minutes of the ECMA Fortran Committee, TC8,

      STANDARDS                meetings of October 8-10 and December 10-12

      ACTIVITIES               1973 were available. Mr van Duin of

Philips-Electrologica and Mr Maisey of ICL

were now chairman and vice-chairman

respectively. Their minutes mentioned

favourably an exchange of documents with

this Group.


TC8 continued to consider in detail the

proposals of the American Standards

Fortran Committee, X3J3, and to communicate

with X3J3. The questions of EXTERNAL &

GENERIC and of input-output were ones of

current interest.


9.    ANSI FORTRAN             Minutes of the American Standards Fortran

      ACTIVITIES               Committee, X3J3, meeting of November 13-16

1973 were considered and relevant pages

are shown in Appendix I to these minutes.


It was pointed out that in allowing real

and double precision DO-loop control

variables Fortran was opposing the trend

in Algol, from Algol 60 to Algol W and

Algal 68.


10.   ROOMS FOR FUTURE         The Society now wished Specialist Groups

      MEETINGS                 not to meet in BCS Headquarters. A

majority of members present expressed

reluctance to move and the Secretary was        DTM

asked to establish whether it was possible

to continue to use the present room.

Barnard's Inn was no longer available as

an alternative.


11.   OTHER BUSINESS           a.    Some ideas were put forward for

publicizing the draft American Standard

when it appeared, including publication

in the Journal, a one-day conference    BHS/DTM

and a two-day workshop. It was

suggested that information be sent to

Computer Weekly so that an article

could appear in "Software File".


b.    A list of about 50 differences between

WATFOR as implemented on the ICL

System 4 and the ANSI Standard and the

IBM Fortran(G) compiler has been

compiled by, and is available from,

Mr Clarke.


12.   DATE OF NEXT             The next meeting will be held on Friday, 15

      MEETING                  March 1974 at 10.30 am at BCS Headquarters,

29 Portland Place, London W.1. It was

decided to hold a meeting after a

relatively short interval because a number

of items, notably Eurocomp and the BSI

representative, need to be resolved

quickly, because Mr Shearing was thought

to have further information on these and

because it was thought that minutes of the

next X3J3 meeting would then be available.




APPENDIX I


[The appendices of the X3J3 minutes were not included in the FSG minutes.]



1.    Opening Remarks


The thirty-eighth meeting of X3J3 was called to order at 10:00 a.m. by

chairman Engel.


2.    Procedural Items


2.1   Attendance


See Appendix B.


2.2   Agenda


See Appendix C for the approved agenda.


2.3   Membership


1.    At the start of the meeting, the membership consisted of twenty-

four members and the chairman. At the start of the meeting,

sixteen voter members were present.


2.    Dave Dillon was welcomed and approved as the alternate member

for Ragan. T.J. Callaghan was approved as a second alternate

member for Erriksson after the membership rules were amended

to allow more than one alternate member (See 4.12).


3.    Weldon is on provisional status at this meeting.


2.4   Previous Minutes - 37th Meeting (X3J3/46)


After several changes and corrections, the minutes of the thirty-

seventh meeting were approved. The outline for voting procedures

was corrected according to information received from Bob Brown at

CBEMA.


3.    Reports


3.1   Chairman's Report


The chairman received a letter from Bob Brown at CBEMA correcting

the outline of voting procedures given to X3J3 at the last meeting.

The minutes of the meeting have been corrected.


Information about TC8 activities and a number of proposals

developed by TC8 have been sent to the voter and alternate members

of X3J3 by TC8 (See Appendix J).


The chairman has prepared an article for Datamation magazine as an

outgrowth of the responses to Thorlin's article on DO-loops.

Other members of X353 are encouraged to submit articles and papers

concerning our work to magazines and conferences.


The chairman has a copy of the responses made to the comments

received during a COBOL letter ballot.


The chairman commented on the persistency of the FORTRAN language

over the years and circulated several documents to illustrate this

point.


The chairman read a note from Stu Lovell that informed us of the

illness and death of his wife. The committee wishes to express

their sympathy and their hope that Stu will be able to rejoin us

at future meetings.


3.2   Editorial Committee


The editorial committee reported that a (73-10-26) version of

FORTREV was mailed last week to the active members and the 29 people

that replied to our survey and requested a copy. This version

contains many changes and has essentially all changes that have been

jointly approved by each group and the editorial committee. The

editor submitted two pages of suggested additional changes for

consideration by the appropriate group at this meeting. During

this meeting, the editorial committee met with Groups 1, 5 and 6.


Due to the small number of approved changes and a request from the

chairman, an updated version of FORTREV will not be distributed

before the January meeting.


3.3   Report of Mail Survey


The secretary reported the following results of the mail survey:


Approx. No. of Letters Sent        98

Replies                            74

Requested FORTREV                  29

Did Not Request Fortrev            45

Address Changes                    19

Requested Removal                   2


3.4   FORTRAN Data Manipulation Language


Greenfield reported that CODASYL has 15-20 people willing to work on

a data manipulation language to interface with FORTRAN. See

Appendix W in X3J3/46.


4.    Actions Taken by the Committee


4.1   Input/Output Clarifications


Gibson moved, Matheny seconded, to adopt item 1 in Appendix G which

allows FMT=* and UNIT=u forms in list-directed I/O statements. The

motion passed (17-0).


Gibson moved, Matheny seconded, to adopt item 3 in Appendix G which

requires that the sequential auxiliary statements designate a unit

connected to a sequential file. The motion passed (17-1).


Gibson moved, Matheny seconded, to adopt item 4 in Appendix G which

would remove the END= specification from direct access READ and

WRITE statements and change all direct access end-of~file conditions

to error conditions. The motion passed (11-8).


Gibson moved, Matheny seconded, that in an "inquire by unit" state-

ment, the logical entity op becomes defined to false if the designated

unit is not connected to a file. This of itself does not cause an

error condition. See Appendix G, item 5. The motion passed

(21-0).


4.2   ENTRY and FUNCTION Statements with No Arguments


Karp moved, Greenfield seconded, to adopt Appendix G of X3J3/46

which would allow the additional form of "ENTRY name ( )" in function

subprograms. Klein suggested that the two forms of an ENTRY state-

ment with no arguments be considered separately. The part of the

proposal that allows the form with the empty parentheses in function

subprograms passed (18~0). The part of the proposal that continues

to allow the form without the empty parentheses passed (l3~?), so the

effect of the two votes is to adopt the proposal as stated in

Appendix G of X3J3/46.


Moore moved, Saunders seconded, to allow the form "FUNCTION name" as

an additional form of the FUNCTION statement. The motion passed

(15-5).


4.3   Mixed Type Arithmetic Expression Evaluation


Ragan moved, Matheny seconded, to adopt the proposal in Appendix I

which states that the evaluation rules for expressions in assignment

statements are not affected by the type of the entity to be assigned.

The motion passed (15-3).


4.4   Rescind Array Cross-sections


Karp moved, Brender seconded, to adopt the ECMA proposal to delete

array cross~sections and array assignment statements. See Appendix

J, pages J6-J7. The motion passed (12-8).


Some arguments offered in favour of deletion were: (1) ECMA is

opposed to the feature, (2) it is not of significant help to an

optimizing compiler, (3) it may inhibit future development of array

operations (See Appendix P in X3J3/45), (4) it is not a new facility,

(5) it may lead to user confusion, and (6) it has overlap problems.


Some arguments offered against the deletion were: (1) it is a

convenience to the programmer, (2) it is a precise way of stating

actions on arrays, (3) it simplifies compiler optimization, and

(4) there are at least three implementations of the feature.


4.5   Restrict Character Associations


Karp moved, Schwebel seconded, to adopt the proposal in Appendix K

which would restrict character associations to entities of equal

length. The motion failed (2-16).


4.6   Rescind Formal Label Assignment


Karp moved, Greenfield seconded, to disallow the use of the ASSIGN

statement to assign format labels to an integer variable. See item

15 in Appendix U of X3J3/46 and page ZC5 in Appendix ZC of X3J3/42.

The motion failed (7-13).


4.7   Integer Control Variables


Karp moved, Holberton seconded, to adopt ECMA proposal 35 in Appendix

J (page J11) which would allow only integer control variables in

DO-loops. The motion failed (4-16).


Some arguments offered in favour of the motion were: (1) ECMA favours

it, (2) real and double arithmetic is imprecise, (3) incrementation

may be inaccurate, and (4) a syntax error check is lost.


Some arguments offered against the motion were: (1) real and double

control variables are useful, and (2) the language is more regular

as is.


4.8   DO-loop Branching


Karp moved, Klein seconded, to adopt the first three sentences of

proposal 4 in Appendix S of X3J3/42 which would require an "extended

range" of a DO-loop to return to the same DO-loop that branched to

that "extended range". The motion failed (1-19).


4.9   Minimum Times Through a DO-loop


Karp moved, Greenfield seconded, that the minimum times through a

DO-loop be one. See item 36 in Appendix U of X3J3/46. The motion

failed (6-15).


The primary argument in favour of the motion was that compilers can

do more optimization easier if the minimum is one. Code that is

invariant within a loop cannot be as effectively taken out of the

loop if it might not be executed at all.


Arguments offered against the motion included: (1) optimization

becomes less important as internal speeds increase, (2) users expect

zero as the minimum, and (3) users can't easily bypass I/O list items

when iteration count is zero.


4.10  GENERIC Statement


Karp moved, Klein seconded, to adopt proposal 3 in Appendix S of

X3J3/42 which would require a GENERIC statement in each program unit

that referenced any generic function. The motion failed (7-12).


Some arguments offered in favour of the motion were: (1) existing

standard programs do not expect generic functions, (2) compilers may

do more argument checking, and (3) there are at least two existing

implementations.


Some arguments against the motion were: (1) except for the external

function selection problem, an existing standard program is not

affected by the lack of a GENERIC statement, (2) it is expected that

users will like generic functions and will consider the requirement

for a GENERIC statement a nuisance and a "millstone around their

necks", and (3) use of generic functions should be encouraged because

they aid in the conversion of programs between single and double

precision.


4.11   End Line


Marshall moved, Senn seconded, to adopt proposal 1 in Appendix L

which would change an "end line" to an "END statement" without any

substantive change in the rules. After some discussion, some argu-

ments about whether the change was editorial or not, and a defense

of X3.9-1966, Klein moved, Greenfield seconded, to table the motion.

The motion to table failed (9-11). After some further discussion,

Bailey moved, Matheny seconded, to table the motion. This motion

to table passed (11-7).


4.12   Membership Rules


Greenfield moved, Gibson seconded, to allow a voter member to have

any number of alternate members. The motion passed (13-3).


4.13  Documents to submit to ISO/TC97/SC5


Karp moved, Noll seconded, to submit X3J3/44, which is the (73-10-26)

version of FORTREV, and an updated version of X3J3/38 that is not to

include items approved at this 38th meeting to ISO/97/SC5 for their

May 1974 meeting. The deadline for submission of such documents is

in December 1973. Noll moved, Saunders seconded to amend the motion

by replacing "X3J3/44, which is the (73-10-26) version of FORTREV"

with "X3J3/44 or the latest version of FORTREV that has been distri-

buted to X3J3". The amendment failed (4-9). The motion passed

(19-0).


5.    Technical Items Discussed But No Action Taken


5.1   Input/Output Clarifications


Gibson moved, Moore seconded, to adopt item 2 in Appendix G. However

the motion was withdrawn when a straw vote (16-4) indicated that the

committee wanted the syntax to include parentheses when keywords are

used. Group 4 is to rewrite the proposal.


Gibson moved, Moore seconded, to adopt item 6 in Appendix G. However,

the motion was withdrawn when the committee indicated that it wanted

a better syntax for identifying the unit or file name that was being

inquired about. Group 4 is to rewrite the proposal.


5.2   Group Reports


Group 1 report: Group 1 met with the editorial committee and agreed

to a number of editorial changes. Changing the concept of an end

line to an END statement was discussed and a proposal written. See

4.11 and Appendix L.


Group 2 report: Group 2 prepared a proposal for a modification to

the rules for the evaluation of expressions in assignment statements.

See 4.3 and Appendix I.


Group 3 report; Group 3 brought a number of proposals before the

full committee. See items 4.2, 4.4, 4.5, 4.6, 4.7, 4.8, 4.9 and

4.10. The group expects to propose the equivalencing of dummy argu-

ments at the next meeting.


Group 4: Group 1 brought a number of input/output proposals before

the full committee. See 4.1 and 5.1. The group did some more

editorial work on section 13 and discussed changes to the proposals

that were not approved by the committee.


Group 5: Group 5 spent its time validating the current FORTREV

document and identifying problem areas. Temporary acting chairman

Engel has dictated that no changes be made to sections 14-16 before

the January meeting.


Group 6: Group 6 met with the editorial committee and discussed some

methods of producing subset documents. However, it was realized

that the committee has not yet approved any specific items for subsets,

so it is early to be producing exact subset text. The group will

prepare proposals on the content of each of the subsets.


6.    Assignments


6.1   Distributor


Material to be distributed before the next meeting should reach

Greenfield by December 26, 1973. Agenda items should be sent to

Bailey.


6.2   Study Of FORTREV


A detailed examination of Section 5 of FORTREV (73-10-26) will be on

the agenda for the next meeting. All members are urged to study

section 5 and prepare written comments before the next meeting.


7.    Future Meetings


Jan 14-17, 1974 Phoenix, AZ (Greenfield (617) 667-3111 X2547)

Week of Mar. 11 Dallas, Holmdel, Hampton Roads, or Washington

Week of May 13

Week of July 8


The January meeting will start at 9:30 a.m. on the first day and adjourn

at 5:00 p.m. on the last day.


8.    Adjournment


The chairman thanked the members and observers for the work accomplished

and adjourned the meeting at 2:45 p.m.


Respectfully submitted,


(Sgd) LLOYD W. CAMPBELL

Secretary

X3J3 FORTRAN