There have been several changes in Group membership during the last quarter. G A England left on 1 February to take up an appointment at Imperial College. P Kent transferred to the Applications Software Group on 1 February and M D Fowler joined the Group from Applications Software on 1 March.
The staff now comprises:-
The code in the 7903 has been stable and operating reasonably well. The line configuration has been altered in order to rationalise the line allocation and to provide for logical expansion. In addition, the multi-drop facility was used successfully to prevent Manchester and Liverpool interaction on the switched service.
The 2400 baud switched modem (PO Model 7C) was delivered and is currently operating at 600 and 1200 bauds. Little effort has been made to operate at 2400 bauds initially because of problems in the Modular One software and latterly due to bad luck in that the 1906A or 7903 has broken down before each of the experiments which had been arranged could take place. Work in this direction will continue.
Negotiations are in hand with RHEL to extend the switchboard to provide lines for the new wing (A4).
Progress has been made in that Imperial College have now discovered that their hardware, or use of hardware, is preventing them recognising the GEORGE 4 sign-on message. The problem now is that time-out on the 7903 is too fast (50 milliseconds) to allow them to respond. A hardware modification is required in the 7903 known as the termi printer mod and this is being negotiated with ICL engineers.
The first 2050 arrived on 7 February after some delay and a few hardware faults have been experienced in the power supply, lineprinter and cassette loader. The fault in the latter was caused by a loose plug and should not recur. The second 2050 arrived on 8 March and is awaiting transport to a site.
Use of the GEC standard software has created problems, revealing insufficient documentation and also lack of paper tapes, in particular the device control code for the lineprinter, card reader and communications equipment. The latter control code is still not available due to GEC's reluctance to release undocumented code.
Familiarity with running most GEC utilities on the machine (as opposed to the simulator on the 360/195) has been achieved and the structure and building of systems is now understood although not achieved due to faults in the code.
Card input should soon be possible with the aid of a simple bootstrap loader which will allow binary card input to the 2050.
In the interests of speed of implementation, it has been decided to construct a 7020 emulator avoiding GEC software where necessary. One hardware problem remains and that is that the communications hardware has been set up to comply with IBM instead of ICL standards and a small hardware modification is needed to rectify this.
The Group's interest in performance has been maintained and various ideas for measurements and experiments have been tried. There have been continual changes to the parameters under which GEORGE runs in order to achieve some optimum. It appears that most parameters are on fairly flat curves and any reasonable value is near optimal. In addition, the character of the workload is so changeable that it is difficult to determine whether performance changes are due to alterations or chance changes in the work profile. Nonetheless, the Group's efforts lead to a greater understanding of the problems and an intelligent and analytical approach to performance improvement. Not least among the factors involved in determining performance is operator activity and the influence, control and informing of operators in various ways is an important activity.
The listing program for :SYSTEM.PERFORMANCE has been substantially modified. A macro to drive this has been produced. The various files associated with the performance have been tidied into directories. The system has been documented. This has led to the automatic production of weekly performance figures which are written to a file for long term analysis.
A close study has begun of the performance in an effort to improve mill efficiency. Some discussions with ICL are in progress.
A paper has been produced on a possible method of costing which could be of great value in controlling the machine. There has been no progress on implementation yet.
A brief study was made and a Technical Notice produced of the store layout of the areas associated with segmentation and paging in GEORGE.
A program has been used to measure the paging and program sharing and a Technical Notice has been produced. This showed sharing of the compilers was rare but that the compilers remained in store for very long periods. Only some 50K of store was used on average. It may be that if more than one program is sharing code, then the measuring program may have been swapped out and produced an incorrect measurement but this is unlikely. The study showed some under-utilisation of the core which subsequent HLS changes have been aimed at rectifying.
The GIVE/6 extracode for reading from the GEORGE area has been used to collect information about the various chains of blocks that GEORGE maintains in its workspace. It is proposed to continue this process under Mk 7 and to draw attention to any significant changes that may be observed especially in connection with file store usage.
It is hoped to incorporate such observations in GERONIMO to give an instant insight into GEORGE's thoughts.
The modification of GEORGE 4 Mk 6 has ceased in the light of the hoped-for imminence of Mk 7. Mk 6 is, however, being updated by ICL mends as required. One of the prime activities of the Group has been the introduction of Mk 7 and the sooner this is achieved, the sooner attention can be re-channelled to other developments.
The reliability of GEORGE has not improved significantly and few, if any, of the bugs encountered have been cured. The usual replies from ICL are that the fault will be corrected in Mk 7 or that it cannot be diagnosed. It appears that GEORGE 4 sites have a higher break rate than GEORGE 3 ones which is disappointing. It is intended, in the light of ICL's performance on bugs, to take a greater interest in persistent bugs under Mk 7. Few, if any, problems have been caused by ACL mends to GEORGE.
The introduction of Mk 7 is being delayed by a bug related to disc transfers which prevents post-mortems being taken. Nonetheless, all the private mends to GEORGE have been fairly easily modified and debugged to run under Mk 7. Only a few minor bugs remain. In addition, mends have been produced (RCP) to provide an automatic news feature for MOP and background jobs which should prove useful. The dump system is being modified to force dumping always to be on to two tapes and always on new ones. Automatic broadcasting before dumping is currently under development (RCP).
Another useful mend is to make CE in background mode (but not foreground) demand fully started context. This is to increase the percentage of jobs which can be saved.
Methods of priority treatment for guaranteed users and of on-line magnetic tape/disc scheduling have been worked out and discussed with Resource Management Branch. A new version of HLS has been produced to prepare for these changes.
The general purpose macro COMP for compilation, consolidation and running the ICL compilers, has been in use to selected users since January and appears to meet their needs. A general release is being held up pending documentation which is in the course of production. Here there is a problem in that COMP has a large number of facilities and documentation must satisfy both the simple and complex needs of users. In all, there are some 30 different parameters which COMP accepts which give greater flexibility economically.
Currently, PLASYD and FORTRAN are supported and other languages will be incorporated later. It is hoped that COMP will fully replace the corresponding ICL macros and provide for virtually all users.
Enhancements are envisaged, particularly in the driving of the consolidator. Here, COMP will supply all the parameters rather than the compiler, thus obviating the tedious FORTRAN program descriptions.
Effort has gone into making CORRECT foolproof. Some anomalies have been identified in some commands which can trap the unwary. In particular there are problems with BREAK INs. The current version is much safer and at worst a user can revert to his old file.
An investigation has been carried out to determine, for direct access files, the connection between the program description lines in a FORTRAN program and the type and qualifier of the file that is actually used. A Technical Notice on this, which also contains other comments regarding the FORTRAN program description lines, is being prepared.
The source code of the GERONIMO driving program has been obtained and after a few corrections it was found to compile into a useable binary. Several enhancements have been undertaken, or are under preparation. The major one of these is to the JOB command. The code for this has been entirely replaced and now distinguishes between saved jobs and others. In addition, MOP jobs and fully-started jobs are indicated. The DECK command now gives the deck numbers of idle decks and a list of outstanding tape requests. The LIST command has been altered to prevent files belonging to :MASTER, :MANAGER or :SYSTEM from being accessed.
An investigation has been made of the quality of t the code produced by the optimising compiler XFEH. The code was found to be good and, in general, well optimised. Such faults as were found were noted and a notice has been circulated.
Little progress has been made here.
This macro, which is used for generating identifier lists for FORTRAN programs has been released for general use. Several user-suggested enhancements have been incorporated.
The META II Translator Writing System has been implemented on the 1906A. This is a simple system which will be used mainly as a teaching tool at Brunel University. However, it is possible that the system could be used for simple preprocessing tasks. It uses a simple top-down syntax analysis algorithm to recognise legal strings in the language. The output is defined by output commands in the META II statements. It consists of certain sections of the original text together with additional text specified by the user. Full details are given in a set of lecture notes produced for a course at Brunel.
Following the decision to issue a preliminary version of SPROGS at the end of this quarter, very little extra design work has been undertaken. However, now that this early version is ready, work will begin on some of the more advanced features such as variables and library facilities.
It is hoped that the first version of SPROGS will be released by the time this report is published. A manual giving the basic facilities available has been written. New facilities will be documented in a file in the GEORGE filestore as they become available.
The following features have been added during the quarter:-
A comprehensive macro is now available in :MACROS to assist in the running of the system. It is noted that use of XFEV rather than XFIH doubles the speed of operation of SPROGS.
A teaching film about SPROGS will be made using the system. This should help debug all the facilities, highlight deficiencies and provide an easy way of demonstrating the abilities of the system.
Work will now start on implementing a library system and provide global and local variables (see previous Report). These two features should be present in the next major release. Other routines to be provided include ellipses, parabolas and hyperbola generation and contour plotting.
Some thought has gone into the provision of the line thickening and different region shapes. Since the SD4020 has no hardware facilities to thicken lines, this has to be done by inserting extra lines on either side of the original one. All these extra lines must be clipped correctly by intervening regions, so there is a considerable overhead in the provision of this feature. It has, therefore, been decided that, if a user knows that clipping will not affect any of these extra lines, he will be able to specify that the extra lines be introduced at the lowest region level, thus effecting a considerable saving.
The shape of a region determines what clipping should be done on lines crossing the region boundary. Currently, all regions are rectangular but work is in hand to provide circular and elliptical regions.
The work on fonts 0] and 02 has now been completed. They are now in the system for users to access.
The upper and lower case fonts, referred to in the last Report as FONT 01 and FONT 02, each contain all the characters within the SPROGS set, the only difference between the two being the upper and lower case alphabets.
At present, a loose-leaf manual is being designed, similar to the one produced for GROATS. The manual will enable the user to see the characters at different sizes and also the layout of each character on the grid.
It is desirable to include a contour plotter package in the SPROGS system. Two existing packages have been obtained; they are the OB07 suite of SD4020 routines and the package produced by the Research Laboratory of the General Electric Company, written in 1965.
A detailed study is being made of these routines and suggestions for improvements to these will no doubt be put forward in due course.
It can take several man years of work to write a contour plotting program. Unfortunately, the Laboratory does not have this amount of time. It is for this reason that the system necessitates the implementation of an existing package.
Some more tracing facilities have been added to the SPROGS system. There is now a spot check tracing mode. This lists out the contents of the SPROGS filestore ie all picture and film files currently in the system, the current region chain and the current display position. Some other routines such as those for opening and closing picture file definitions also output tracing information if any is requested by the user.
Some more routines have been written for use only with the SD4020. These allow the user to request the retention of his magnetic tape, to request that the 35mm camera be used for microfilm and to use the expand/reduce and forms flash features of the SD4020. The line drawing routine has now been improved so that the ends of contiguous lines now meet perfectly.
A modified paper tape reader handler was produced. It reads paper tapes generated by the 1906A. It was obtained from a standard handler by making line-feed act as a line terminator instead of being ignored (SRPOGS Technical Paper 8).
PB is taking over some of the responsibility for the operation and maintenance of the PDP15. To this end, a document describing the VIEW program and relevant parts of the DOS has been produced for use by the Operations Branch (SPROGS Technical Paper 9). The VIEW program will then be usable by people who would not otherwise have access to the PDP15.
If, as now seems possible, the RSX Plus operating system code will be available to us next quarter, some work will be necessary to convert the device handlers and low-level routines which were written or modified here. The device handlers include a significant amount of code which interfaces with the operating system, and which will therefore be used to be rewritten.
JRG and RCP took over the BSI (PDP15/1906A) at the end of January when GAE left. The rest of January was mainly concerned with learning about the 2 devices and understanding the programs that had been written at that stage. RCP was to deal with the 1906A and JRG with the PDP15.
At that stage a set of programs and GEORGE macros had been written which transferred a file from the PDP15 to the GEORGE filestore to be listed on the lineprinter. There was one frequent bug which still persists. It does not occur on every transfer and bulky files have been sent across without incident. Certain crucial points in both programs have been altered without any effect. At the time of writing (16 March) two DEC engineers are due to look at the PDP15 end, although it is not known whether or not this is where the fault lies.
Programs for sending messages in one-direction only, function without error.
A method of listing PDP15 source files on the 1906A lineprinter is now available. The program, LISTIT, (see SPROGS Technical Paper 10) on the PDP15 and the GEORGE macro, PDPLIST, communicate character information via magnetic tape.
Keyboard input and argument-passing routines mentioned in the last report are documented and in use by Brian Woodward in his architectural design programs. Those completed in the last quarter as part of PIGS include:-
Although a good deal of work remains to complete the PIGS system, the set of routines currently available should greatly reduce the graphics expertise necessary to write an interactive program. Prospective users are encouraged to consult WDS.
The EMS VCS3 synthesizer has arrived and has been checked. The logic modules and associated racking have been ordered from DEC and most of these parts have now arrived. A wiring diagram and list of connections for the back plane wiring have been made. This and the necessary components will be sent to the Rutherford Laboratory workshops when the remaining components arrive from DEC.
A new GROATS macro has been written and installed in :MACROS. This macro has standard parameter formats in that it looks (to the user) like any other ACL GEORGE macro. Chapter 6 of the GROATS manual had to be rewritten to cater for this. The old macros are still available to users, but will eventually be withdrawn.
There has been some work carried out on improving the speed of the system, and to this end the Algol routines CONVERT, DRAWVECTOR and RASTERLENGTH have been rewritten in PLAN. On various test jobs, this has resulted in reductions in mill time of from 18 to 29 per cent.
The new manual for CAMP and CAMPER has now been approved and is being printed. Some minor modifications have been made to the programs in order to remove some errors from the original systems.
Most of the quarter has been spent in finding out the reasons for the relative movement between colours in multi-colour prints. Bad registration in the SD4020 camera was eliminated by some tests early in the quarter. A small amount of movement was detected in the position of the image on the SD4020 tube but this is not sufficient to account for the errors found in the prints. The conclusions reached eventually were that the tolerances allowed in the British Standard on 16mm film stock are such that the errors originate in the printing process and that it will be impossible to get rid of these.
It will, therefore, be necessary to generate multi-colour prints by using 35mm film initially and reducing the final colour print to 16mm as the last stage in the process. The cost of this process is significantly greater than using 16mm throughout and it may be difficult to justify it except in special cases. There is difficulty currently in finding a film laboratory willing to process our Dacomatic film in quantities greater than 100 feet. The tests done so far suggest that the movement will be greatly reduced but not completely eliminated.
Due to the problems outlined above, the decision was made to produce the final versions of the Hash films as white-on-blue copies. Three of the set of five have now been completed and work is progressing on the other two.
The system has been polished to some extent and documentation on using it has been written. The game was used for an SRC Management Training Course in March and, apart from trouble with the 7903 on the first evening, it ran successfully: even GEORGE stood up without a break!
Work is well progressed on a new version of the control program which will enable the controller to communicate with the system while it is running programs for the players. This new version will also be able to recover from losing one or more of the players without having to crash the (COLAB) system and re-starting it.
When GEORGE Mark 7 becomes established, it is hoped to convert the system from using communication files to using a conceptual multiplexor; however, a certain amount of information will be required to see if this is possible.
With the introduction of SPROGS, the more advanced features of it will be implemented as indicated earlier. Undoubtedly maintenance of SPROGS will absorb much effort.
On the 1906A front. the major event should be the introduction of Mark 7 and its exploitation. If the current lack of progress by ICL in curing bugs persists. it may well not be introduced during the next quarter.
A great deal of effort is likely to be expended on improving the efficiency of the 1906A in various ways and some user-visible enhancements will also be made.
During the next quarter. a preliminary version of the 7020 emulator on the GEC 2050 should be available. It is also hoped that during this time more definite plans as to the future of communications at the Laboratory will be made. To some extent this will depend on plans for the next large machine. Negotiations are in hand with the Post Office to use the new packet switching service in a few years' time. The GEC 2050, Sigma 2, a new front end for the 1906A, or the new large machine could use this service, depending on what use ACL can make of it.
The Quadratic Hash when the table size is a power of 2, Computer Journal, Nov 1972, F R A Hopgood
A set of 9 lecture notes for a fourth year course at Brunel University on Translator Writing Systems:
COLAB Computer Users' Guide, D C Toll
COLAB Controllers' Guide to Computerised COLAB on the 1906A (Basic Peripheral Communication File System), D C Toll
Notes on GROATS for use with GEORGE 2, D C Toll
1906A New GROATS Macro, D C Toll
PDP15 Routine Library, J R Gallop and W D Shaw
PDP15 Routine Library Update No 1, W D Shaw
Lecture Courses on Compiler Writing Techniques, Translator Writing Systems and Computer Animation given at Brunel University during the whole of the Spring Term (16 January to 20 March), F R A Hopgood
19 January, Talk on GEORGE 4 - Use and Experience at ICL 1900 Series Operating Systems Seminar for consultants, P E Bryant
27 February, Talk on Modifying GEORGE at a GEORGE 3 Colloquium at RSRS Slough, P E Bryant
ACTP Software Committee (2 meetings), Various visits and papers on behalf of ACTP, GEORGE 3 User Group General Meeting, London 1 February, P E Bryant
ICLCUA Performance Activity Group 15 February, G W Robinson
ICLCUA Remote Users Activity Group 16 February, F V Sims
University GEORGE User Group, Nottingham 1 March, P E Bryant
ICL New Range Working Party (2 meetings), C J Pavelin
PDP15-Local Users Group Meeting 20 February, W D Shaw
