MINUTES OF BCS FORTRAN SPECIALIST GROUP MEETING HELD AT BCS HQ

                          ON 1ST APRIL 1987

Present:        Mike Bell            - Middlesex Polytechnic

                Mike Bennett         - CEGB Training College

                Duncan Browne        - Watson Calculating Services

                Francis Cox          - Swifte Computer Systems

                Ray Crawford         - RAE Pyestock

                Shah Datardina       - NAG

                Miles Ellis          - Oxford University CTC

                Bill Flexner         - retired

                Dave Griffiths       - SSF

                Carol Hewlett        - LSE

                Peter Holland        - SSL

                Chris Lazou          - ULCC

                Clive Massey         - SWURCC

                Brian Meek           - Kings College, London

                Mike Nunn            - CCTA

                Denis Parkinson      - QMC

                T. van Raalte        - MOD (PE)

                John Stratford       - QMC Computer Centre

                Tony Webster         - Salford University

                Alan Wilson          - ICL/AMT

                John Wilson          - Leicester University

                John Young           - MOD (PE)

l.   APOLOGIES FOR ABSENCE

Apologies were received from Mike Geary, Alan Jenyon, Heather

Liddell, David Muxworthy, John Reid and Lawrie Schonfelder.

2.   MINUTES OF PREVIOUS MEETING [8 January 1987]

Correction to p6, final paragraph:

.....the UK voted "Yes with comments" to releasing the 8X

draft for public comment.

3.   MATTERS ARISING

Brian Meek mentioned that BSI is already receiving reports

from various people on the "8X" draft. When the final version

is released by X3J3 for public review, the BSI Fortran panel

will be reconvened to formulate a UK response and our Group

will be expected to provide an authoritative view.

4.   BCS BUSINESS

     (i)    BCS is launching a publicity campaign to encourage

            people to apply for the Chartered Engineer status - see

            March edition of "Computing".

    (ii)    The Software Engineering Technical Committee of BCS

            recently invited our Group to send a representative

            its meetings. So far our chairman, John Wilson, has

            taken on this role but can no longer afford the time.

            Therefore we need a volunteer to take this function

            over - please note that expenses incurred in attending

            are reimbursed by BCS. John would like to hear from

            anyone interested in becoming our representative and

            can supply more information about what is involved.

   (iii)    Brian Meek drew attention to a joint BSI/BCS Group

            which meets from time to time to discuss activities

            Standards. He himself is a member.

    (iv)    BCS HQ is drawing up a list of nominations for

            distinguished fellows. Anyone in the Fortran Group

            with suggestions should pass them forward.

     (v)    Suggestions for talks at future meetings of our Group

            included:

                 Fortran on small parallel processors

                 Fortran 8X compared with Ada

5.   FORTRAN FORUM

Alan Wilson reported that editorial work on the 8X draft was

completed at the last meeting of X3J3 and a vote will take

place to decide on its readiness for submission to the X3

parent committee prior to a release for public review. According to

Miles Ellis, Jeanne Adams believes it may even be possible to release

the draft for public comment in parallel with its consideration by X3.

Our Group will arrange to get copies which will be purchasable as part

of the booking fee for those attending the Forum. It will probably also

be possible to offer a summary document for those preferring it. The

date for the Forum will be established when the release date for

the 8X draft is known. It will be organised by a subcommittee

including Chris Lazou, John Wilson, John Reid and Miles Ellis.

6.   AGM

  (i)   The following officers were nominated for 1987/8 and

        elected unopposed:

                Chairman             - John Wilson

                Vice-chairman        - Chris Lazou

                Secretary            - Mike Nunn

                Treasurer            - Miles Ellis

 (ii)   In the absence of the previous Treasurer, John Dyke, no

        statement of accounts was available at the meeting.

        However, John has now put the necessary statistics

        together - see appendix A.

7.   PROGRESS ON FORTRAN 8X AT X3J3

 (i)    Miles Ellis reported that there had been a letter ballot

        of X3J3 members over Christmas on the suitability of the

        8X draft. Thousands of comments were received and all

        had to be processed. A sub-group of X3J3 including

        Jeanne Adams met at Albuquerque in January to work

        through them and put them into categories eg "editorial",

        "matters of great substance, whether in, or out" etc..

        So, at the later Los Angeles meeting an indexed

        list of comments was available with their relative

        importance signified. The LA meeting mainly concerned

        itself with editorial work; only one significant technical

        change was made. Miles himself proposed that there

        should be something in the Standard about processors

        being required to report deviations from the language as

        specified by Fortran 8X syntax laws. Jeanne Adams was

        also in favour and this idea was incorporated.

 (ii)   "No" votes on the suitability of the 8X draft were

        registered by DEC, Unisys, Cray, IBM, Boeing & Peritus.

        Reasons included:

        IBM-    thought there should be two languages, viz (i)

                F77, plus a few extensions. (ii) everything else

                which should be processed in some other way (e.g.

                a Journal of Development).

        DEC -   want to take some things out.

        CRAY -  objections were either editorial or where they

                saw technical holes.

        Because there was so much work to do in LA an extra

        meeting was convened for Albuquerque, where Alan Wilson

        said that 3500 edits were completed. A tremendous job

        had been done in this respect by Lloyd Campbell who had

        also been responsible for editing the F77 standard.

        Walt Brainerd arranged for the document to be printed

        before the following meeting in Seattle. John Reid's

        reports on the latest X3J3 meetings appear in Appendix B.

(iii)   Our X3J3 representatives were asked why some suppliers

        were against the 8X draft. Miles Ellis said that IBM

        believed most users would not want the new features and

        the fact that a more complete language might run slower

        was also unattractive to them. DEC was possibly against

        it because it had already invested a great deal of

        effort into F77 extensions in its current Fortran

        implementation.

 (iv)   Ray Crawford (RAE) asked about security features in the

        8X Standard. Alan Wilson and Miles Ellis said there was

        nothing because this was a function of the operating

        system rather than a compiler.

8.   ANY OTHER BUSINESS

(i)     NCC had provided some documentary materials on how the

        compiler validation services for Fortran and Cobol

        operate.

(ii)    John Wilson had given a talk on Fortran 8X to the

        previous week's DECUS meeting in York. Response to the

        language likely to emerge as 8X had been generally

        favourable.

9.   TALK - "ACTIVE MEMORY TECHNOLOGY DAP-3"

At the afternoon session Professor Denis Parkinson (Queen Mary

College and Active Memory Technology Ltd.) gave a talk on the

AMT DAP-3. A summary appears in Appendix C.

10.   DATE OF NEXT MEETING

The next meeting of the Group will take place on Thursday 2nd

July between 11am and 4pm at Coventry Polytechnic - see map

and directions attached. In the afternoon, starting at

1.45pm, there will be a talk on "Productivity Tools for

Fortran Programmers" by John Appleyard of Production Software.

ll.   MISCELLANEOUS

One of our Group, Mike Metcalf of CERN, has written a book on

Fortran 77 which is described in Appendix D.

MIKE NUNN

Secretary, BCS Fortran Specialist Group

18th May, 1987.

                            APPENDIX A

                     THE BRITISH COMPUTER SOCIETY

                  13 Mansfield Street, London w1M 0BP

                    RECEIPTS AND PAYMENTS ACCOUNT FORM

                        Fortran Specialist Group

Receipts & Payments Account for the year ended 30th April, 1987

(Please submit to HQ by 25th May)

BCS Fortran Specialist Group - Income and Expenditure - as at 30.4.87

                                    APPENDIX B

To:        ISO, Fortran Forum, BCS, NAG, etc.

From:      John Reid

Date:      30 March 1987

Subject:   X3J3 meeting in Albuquerque

Note:      This is a personal report of the meeting and in no sense

           does it constitute an official record of it.

1.   Summary

     This was technically a continuation of the February

meeting, called because of the volume of comments received

with the ballots. About half the members were present,

including almost all those who suggested large numbers of

changes. It was a very hard-working and productive meeting

and at its end everyone seemed satisfied that the comments

had been considered adequately. The plan is to incorporate

all the accepted changes into the draft Standard and vote on

the resulting document at the beginning of the May meeting.

I am expecting to find it ready for release for public

comment and am hoping that the necessary two-thirds of the

Committee will too. We need also to respond formally to all

those who submitted comments. Constructing and approving

these responses will be a significant task for the May

meeting.

     The bulk of the changes were editorial and the details

are not interesting to readers of this report. We did,

however, make a change to the default implicit rules

(Section 2), that I find very welcome; change the syntax of

DO ( ) TIMES (Section 3), again a change that I welcome;

and there were various minor technical changes and

clarifications (Section 4).

2.   The default IMPLICIT rules

     The default IMPLICIT rules were complicated and several

members commented adversely in their ballots. I was not one

of them, but was keen to see a change, so I participated in

a small working party and co-authored a paper during the

meeting. We proposed that in a nested set of scoping units,

each should take for its default the IMPLICIT rules of its

host, with the Fortran 77 default in the outermost one. In

each scoping unit, IMPLICIT rules might be used to override

the defaults for any or all of the letters. Safe practice

is to use IMPLICIT NONE, particularly in inner scopes, but

we felt that it is wrong that sometimes this should be the

default and sometimes the Fortran 77 rules should be the

default. Also it should be possible to define the IMPLICIT

rules throughout the whole of a program unit by a single

statement at its head. We were encouraged by very

favourable straw votes and the proposal passed (15-0).

3.   The syntax of the DO statement

     Larry Rolison suggested that the syntax of the statement

                        DO (scalar-int-expr TIMES)

be replaced by

                        DO (scalar-int-expr) TIMES

in order to make parsing easier. The Committee was rather

undecided about whether such a change was really necessary,

but eventually approved it (10-5). I have been asked for

this change, and supported it.

4.   Miscellaneous minor changes and clarifications

     It was decided to allow a specification expression to refer to

the result of another specification expression, for example

                        INTEGER B(2*SIZE(A))

only if the expression referenced comes earlier in the

specification sequence. This disallows such a case as

                        INTEGER A(2*SIZE(B)), B(2*SIZE(A))

     It was agreed that constant arrays may have the RANGE attribute.

     It was decided to allow construct names to be used on

ELSE IF, ELSE, and CASE statements.

     Automatic arrays are arrays that are not dummy

arguments but have variable extents. It was decided to use

the term "automatic object" for both these and structures

with varying type parameters since these have similar

properties. For example, matrix arithmetic might be

implemented by a type MATRIX that had parameters for the

numbers of rows and columns. Automatic objects are created

on entry to a procedure and lost on return.

     It was decided (8-5) not to permit statement functions

in modules.

     Wording was accepted (13-0) to clarify the rules that

an alias may have only one parent. It is acceptable for an

alias to be associated with two subobjects of the same named

object.

5.   Next meeting of X3J3

     The next meeting of X3J3 will be in Seattle, 11-15 May.

The premeeting deadline is April 6.

[the following report was written after the Fortran SG meeting of

1 April 1987 but was included with the minutes of that meeting.]

To:        Fortran Forum, BCS, NAG, etc.

From:      John Reid

Subject:   X3J3 meeting in Seattle

Date:      18 May 1987

Note:      This is a personal report of the meeting and in no sense does

           it constitute an official record of it.

1.   Summary

     The meeting was devoted to deciding whether to send the

current draft standard to X3, the parent committee, and

processing the responses to the comments made by members

with their letter ballots on this issue. The decision was

"yes", but the vote (26-9) was such that there is likely to

be a delay before the draft is released for public comment

while X3 considers the comments accompanying the "no" votes.

This was a disappointment to those of us who wish to see the

public comment period beginning as soon as possible, but at

least we have reached another formal milestone.

     Some editorial changes were made (few compared with the

previous meeting) and one technical change was made (see

section 6). It was a tense meeting and there was some

discussion of the membership rules, but the friendly

personal relationships that are a hallmark of X3J3 were

maintained.

2.   X3J3 secretary

     Jeanne Martin has resigned as secretary and I have been

appointed in her place. She has performed her duties very

competently and conscientiously since 1982 and was warmly

applauded by the committee to mark its appreciation. I

thought hard about whether to accept and decided to do so

only on conditions that will greatly reduce the burden and

that I thought might be unacceptable because the quality of

the minutes will deteriorate. I am recording formal

decisions only and relying on committee members to provide

scribe notes that I will use to construct the minutes by

"cutting and sticking". Since I always have recorded the

committee decisions for myself, this will make little

difference to my participation in technical discussions.

     I would like to make it clear that as secretary I will

do my best to ensure that the minutes are an accurate and

unbiased record of the meeting, but these notes are personal

notes and do include my personal opinions. They are

constructed by making minor changes to a report that I write

for people in my own organisation to explain how I have been

representing them.

3.    Voting during the public comment period

     The chairman has received a ruling from X3 on changing

the draft standard during the public comment period. A new

standing document (S12) will be created, starting with the

present draft standard. Previously, it had been assumed

that changes to S12 could be made by normal majority votes

and that the 2/3 rule would apply only to the final adoption

of S12 as a replacement for S8. However, we have been told

that a 2/3 vote is required for all changes to S12.

Personally, I approve of the rule; if about half the

committee think one way and the other half think the other,

I think that we should keep what we have. However, the

immediate effect is that those that do not like aspects of

the present draft are concerned that it will be harder to

make future changes. This was a reason for several "no"

votes.

4.   The ballot on forwarding the draft

     The vote in the January letter ballot was 29-7 in

favour. The changes since then were as follows: Dick

Hendrickson (Cray) switched from "no" to "yes", being

satisfied with the responses to his comments. One member

who voted "yes" has left the committee. Three members were

absent (two "yes" votes and one "no" vote in the letter

ballot). There were three new members (two "yes" and one

"no" vote). Three members switched from "yes" to "no".

Thus the final vote was 26-9 in favour. The following voted "no":

Steve Adamcyzk (Adv. Comp. Tech.) voted "yes" in the letter

                ballot, saying "I do not agree with everything in

                the draft, but it is a reasonable compromise and

                is ready to go out for public comment". He now

                thinks that it should be limited to

                standardizing existing practice and should

                contain more explanatory text to assist public comment.

Bob Allison (Harris) voted "yes" in the letter ballot,

                saying that he had serious misgivings but was

                willing to see if a public review indicates that

                the public has similar misgivings. He was

                surprised by the large number of comments and

                worried by the clarification of the 2/3 majority

                rule, which will make it more difficult to change

                the draft during the public comment period.

Valerie Bowe (Unisys) reaffirmed her letter ballot.

Kevin Harris (DEC) reaffirmed his letter ballot.

Alex Marusak (Los Alamos) voted "yes" in the letter ballot

                but changed to "no" because the clarification of

                the 2/3 rule will make it harder to change the

                draft. In particular, he wants to add BIT type.

_

Len Moss (SLAC) was represented by Sunny Sund, who reaffirmed his

                vote and was concerned about the 2/3 rule.

Ivor Phillips (Boeing) reaffirmed his "no" vote.

Alison Wearing (NCC, Manchester) is a new member who feels that the

                document is not sufficiently well written.

Dick Weaver (IBM) reaffirmed his "no" vote.

5.    Membership rules

      There was some discussion of the membership rules.

Familiarity with the whole language will become very

important during the public comment period. In fact, a new

member has only to attend one meeting on provisional status

and is thus given preference over an old member who reverts

to provisional status if he or she has not attended two of

the previous three meetings. There was sentiment at least

to apply to the same rule to new members as old ones, but

the standing rules may not allow this.

6.    Changes to SET RANGE

      At the request of implementors, it was decided that an

array with the attribute RANGE should have no effective

shape until a SET RANGE statement is executed for it.

7.   Interpretation of the LEN function in Fortran 77

     The committee received a request from Andrew Tait of

Amdahl for an interpretation of permissible arguments for

the LEN function. He was told that in a standard-conforming

program, the argument must be an expression and not an array

or procedure name.

8.   Multibyte character sets

     Akio Aoyama and Hideo Wada from Japan made a request

for X3J3 to consider character sets, such as Kanji, that

need a two-byte representation. One possibility might be to

introduce another type parameter for type CHARACTER, with a

default value that corresponds to the usual one-byte

representation. X3J3 pointed out that characters may

presently be implemented in more than one byte, but this

would probably be too inefficient. Another possibility is

to set up an intrinsic module, though i/o may be awkward.

The committee decided to explore this further as an example

of the use of modules.

9.   Presentation by David Jones of Microsoft

     Microsoft has its headquarters nearby, and David Jones

requested to make a short presentation. He was satisfied

with the quality of the document and estimated that Fortran

8x is about double the size of Fortran 77. In a user

survey, he found little interest in the Fortran 8x features

but a demand for common mainframe extensions.

10.  Date of next meeting

     The next X3J3 meeting will be in Liverpool, August

10-14. Contact J.L. Schonfelder, Computer Laboratory,

University of Liverpool, Liverpool L69 3BX, if you wish to

attend. ISO/TC97/SC22/WG5 is also meeting in Liverpool,

August 3-7. Again, contact J.L. Schonfelder for details.

                         APPENDIX C

            TALK ON "THE ACTIVE MEMORY TECHNOLOGY DAP-3" by

                 Professor Denis Parkinson (QMC/AMT)

Background

Active Memory Technology is a totally new company set up recently to

exploit the ICL DAP. It could be considered a "management buyout" from

ICL. Most of its funding is venture capital. ICL itself only holds a

minority of AMT shares.  Patents on DAP are now owned by AMT. (ICL

will remain as distributor for DAP in the defence market place).

AMT staff currently number 22 but are expected to 60 by the end of 1987.

Products

Two product ranges are being developed - the DAP 500 and 700 series

[Nb "5" stands for 2**5 i.e. 32 x 32 and "7" for 2**7 i.e. 128 X 128].

Specifications will be:

        DAP 510-4                4Mb memory        32x32           l0mHz

            510-8                8Mb memory                        l0mHz

            510-32              32Mb memory                        5mHz

        DAP 700 series                            128x128

Beta testing of DAP 500 will take place at QMC this Autumn then elsewhere

in early 1988.

AMT DAP is an attached processor to a host running UNIX.  First hosts will

be Microvaxs and SUN workstations. Channel connection allows real-time

online graphics capability.

Uses

AMT DAP offers very high powered interactive computing power on the desk -

equivalent, say, to 0.25xCRAYl or l000xVAX 11/780. But cost is only £100k

for 32x32 DAP + 4 Mb of memory.

State of Development of AMT DAP

A machine is running in an office environment at QMC and four others

are installed at Reading. Compared with the old ICL DAP, DAP-3 uses

new chips housing 64 processors. It also offers greater reliability by

doubling up the processing elements to check results by duplicating

all calculations i.e. processing elements are effectively 32x32x2.

Software

Fortran               - AMT offer Fortran Plus which is

                        Fortran77 with array facilities.

Subroutine Library    - a set of high level utility routines

                        for aiding program development is offered.

Runtime support/      - tools are offered to permit interactive

        debugging       examination of DAP's state on error conditions.

                      - typical source level debugging aids are available.

Simulator             - a crosscompiler which can run on several hosts

                        produces code in development mode to simulate DAP.

All the above software is now running on SUNs and PERQs. Manuals are

currently in production.

Miscellaneous

Professor Parkinson said that the ICL DAP had been ahead of its time,

but not enough funds were available to fully exploit it. One

limitation was that it had to be attached to large 2900

mainframes. However, a large potential market for a DAP-type machine

was with attachment to hosts such as VAXs, but ICL was not in a

position to exploit this so an initiative had to come from outside the

company. ICL also felt that DAP was not to the forefront of its

mainstream efforts, but was to be congratulated in that, although it

was not in a position to further exploit it itself, it was happy to

help a new venture like AMT to do so.

Appendix D

For experienced programmers who want a faster and more complete approach

to FORTRAN 77

•  A practical and up-to-date guide to FORTRAN 77

•  The only available book at this level

•  Written by a central figure in the world of FORTRAN

      programming

At present FORTRAN 77 is the only practical computer language for large-scale

scientific, numerical, and engineering applications. This book provides a concise

but complete description of FORTRAN 77 firmly based on the standard, and

practical advice on program portability, style, design, efficiency, testing, and

documentation. lt is ideal for those who work in the real world of FORTRAN

programming.