The only alteration to Group membership has been the return of Anton Walter on 1 April after 18 months with Applications Software. Further interview boards have been held and, as a result, three new people will be joining the Group in the autumn.
The present structure is:
In the words of the schoolmaster, the term has shown steady progress. No major products have been launched. On the other hand, there has been good progress in all areas. In particular, the SPROGS documentation has been issued and the NUTS system is in listed trials.
The Group has welcomed the services of Chris Hims and Alan Pitchers, both sandwich students who will be leaving to return to college shortly. Dennis Parkinson, on loan from ICL for 8 weeks during this quarter, also made a worthwhile contribution to the Group effort.
The production of a 1906A Reference Manual has been undertaken which is proving a mammoth task, but is considered well worth the effort. Most members of the Group have been involved in producing the first draft, all Parts of which have now been completed. It is hoped to complete the manual late in the autumn.
Since GEORGE 4 Mark 8 has not yet emerged, progress during the quarter has not been as expected. Unfortunately, it should be delivered in a week or two - 'unfortunate' since the arrival of the GEC 4080 and holidays will preclude any concentrated effort on implementing the system.
It had originally been decided to shelve any further GEORGE developments pending Mark 8, but in the event it has been necessary to continue development of Mark 7, albeit at a fairly low level.
It is understood that the editing scheme for updating GEORGE Mark 8 has been the stumbling block. ACL hopes that this system will simplify further GEORGE enhancements.
A Mark 8 microfilm has been produced (PEB) and an index (CJP). Many of the ACL mends have been examine with a view to Mark 8 implementation. An up-to-date version of Mark 8 code is now to hand and a further microfilm and index will be produced.
There has been slow progress on the front end processor project. This is mainly because the staff involved have been unable to divert themselves from their other tasks. With the delivery of the hardware (early July), progress should be considerable in the next quarter.
There has been a steady stream of mends from ICL amounting to some 30 in number. GEORGE has been kept well up to date. The break rate due to software errors has improved with only one major unsolved bug now extant. It appears that a break rate of 1 or 2 a week should be possible in the near future. It is unfortunate that hardware reliability has not improved to anything like the same extent. There is a school of thought that suggests that the inherent reliability of the 1906A is not high and the current hardware break rate is unlikely to be substantially improved. Itemised below are the ACL changes to GEORGE.
Backing store unjammer (CJP). This was changed two years ago and has become less effective and sometimes DUMPER was being called when there were still on-line files unaccessed for months on the discs. In addition, the filestore statistics produced by the unjammer since Mark 7 was introduced were found to be in serious error. The new strategy appears to give an acceptable performance. Statistics are now sent to JOURNAL for analysis (SRP). The analysis forms part of the ACCOUNTS run.
Dr Parkinson (ICL) produced a simulator for some of the functions of the unjammer. This gives some unexpected results which will probably be used for a review of the strategy in Mark 8. No further changes are anticipated in Mark 7 as the mend area is full.
Efficiency (CJP,CH). A jump in efficiency to around 50% in object programs has been maintained for several weeks and this appeared to coincide with the removal of a diagnostic mend to Executive. This is still not proved but the benchmark project should provide the proof.
This event, and the arrival of the ICL 'DATAPASS system (which collects GEORGE performance data and provides means of controlled benchmarking) stimulated the continuation of some earlier work on benchmarks and job costing.
The benchmark system is designed to measure changes in performance due to changes in GEORGE, Executive, or basic systems. The measurements should be fairly accurate, and certainly more accurate than hitherto.
A preliminary series of benchmark tests is currently being carried out with the aim of comparing the efficiency of various versions of Executive.
DATAPASS is crucial in providing the measuring system for the benchmark.
SYSTEM JOURNAL (CJP, CH, PEB, SRP, MCB, AMW). Some errors in the JOURNAL interface package have been found and corrected. The interface is being amended to access the compacted JOURNAL if necessary, and to take advantage of the format of JOURNAL archive tapes written by PEB's program instead of COPYOUT. This program is being amalgamated with MCB's program. The current COPYOUT system has worked fairly well with only one error in the index which led to an invalid set of job statistics which had to be re-run. :NEWS.JOURNAL ALBUM contains brief documentation of the current JOURNAL utilities.
A utility to give details of magnetic tape and deck failures in the form required by Operations Branch is being produced by AMW. The raw information is in JOURNAL.
A compacting program is being written (CH) to condense JOURNAL. This replaces all the entries concerning one job by one composite entry. A scheme whereby the operators can initiate compacting and archiving has been written and is being implemented. In addition, the past JOURNAL files since September last are being compacted by several overnight runs.
The High Level Scheduler now outputs statistics to :SYSTEM.JOURNAL for every express job to help keep track of the commands used. The use of :SYSTEM.PERFORMANCE for collecting HLS statistics has been discontinued.
Progress on further re-structuring of the accounting and budgeting programs has been severely limited in this quarter by a sudden spate of faults in the original central accounting scheme. The spurious overwriting of fields not associated with accounting has occurred. The mending of the database has been time-consuming. The faults themselves have not yet been found despite extensive effort.
The new filestore accounting scheme has been running for some time and information on the weekly situation has been produced and passed to Operations Branch. The filestore size has increased considerably while the control system has awaited implementation.
Forms and output are now compatible and there are no further obstacles to control. A two-page summary of filestore usage of on-line and off-line files, retrieval and unjamming details has been produced. Statistics show that the on-line capacity is accessed once every 5 days as opposed to once every 10 days some weeks ago.
The accounts now use the new security scheme which allows the system to be run without risk to security.
The directory :ACCOUNTS has been moved from being under :MANAGER to a low level in an attempt to reduce the size and complexity of high level directories.
A list of usernames which had not been used during the last financial year was produced and passed to Resource Management Branch for possible action.
Some assistance is being given to Resource Management in rationalising and documenting the large number of database utilities. A bug in the database interface has proved elusive.
Although little development work has been undertaken, considerable effort has been needed to maintain the system. This has been occasioned by the frequent new versions of compilers. The most important enhancement has been the use of the block read facility for reading source into compilers. This is complete for PLASYD and the FORTRAN version is imminent.
Some tidying up has been achieved, particularly in the area of some of the more obvious error halts that occur and on the TASK/compiler interface. The ability to RUN exofile binaries has proved useful.
In order to save filestore space temporarily, TASK creates named semi-compiled files of 5K instead of the normal 10K. Thus small files are not so wasteful but as the file is increased in 10K steps, medium size files may suffer. The TRIM facility to trim a semi-compiled file to minimum size should completely eliminate the problem.
The TASK statistics are about to be analysed by SRP. The raw figures will be inserted in :SYSTEM JOURNAL. An investigation into the efficiency of TASK is planned for the next quarter. This is important as it could benefit a large proportion of the jobs on the 1906A.
NUTS has finally emerged from its shell! The kernel has had all the general purpose parameters implemented (eg ER, EJ, JT, MON, MZ, NOLIST, PR etc) and the basic framework for inclusion of utilities now exists. Currently lacking are routines for magnetic and paper tape handling.
The kernel analyses the parameters and loads an appropriate utility overlay. The utility accesses the parameter information and accesses the required files via kernel routines where error handling is undertaken.
The large variety of utilities and parameter formats has caused problems in that the generality of the system means that little relaxation in the syntax of parameters is currently possible and error checking is difficult. A tiresome aspect is the need to use peripheral types on most filenames to avoid confusion with NUTS parameters.
Utilities which are currently available are:
NUTS is now in limited trials and should be announced in the next quarter.
The source of the PLASYD compiler has been obtained from QMC and work has centred on removing bugs, providing more facilities and improving the efficiency of the compiler. The version 3F contains the features.
A further version will include an AMED type disc editor. This will conclude development.
A file named SUBGROUPS-PL has been created (AJHW) on the libraries disc 26. This is now the default library used by the PLASYD compiler and contains the JLIB routines plus MONITOR, MONITORX and MONITORI from the PLAN library. Further routines will be included as requested and/or supplied.
Long error-free runs of GEORGE have revealed two bugs which have now been cured. Reliability is now excellent. Some effort may be needed with the introduction of the Secure Jobwell.
Some time has been spent looking at semi-compiled to see if any significant improvements can be made. A proposal has been put forward for a new form of intermediate program, relocatable binary, and thought has been given as to how it could be implemented. It is not clear that alterations to the whole compiling system of this magnitude are justified considering the small amount of time usually spent consolidating, especially as compatibility in operation with normal semi-compiled seems desirable. Statistics from the TASK system will be examined to help decide this question. On the other hand, thought has also been given to producing an ACL consolidator similar to XPCH but with extended facilities (eg, overlays, dynamic routine loading, program chaining) and a better interface to host systems such as TASK. This latter alternative currently seems preferable.
A FORTRAN-to-FORTRAN optimising compiler (as described in The Computer Journal) has been obtained in source form. Currently, it is being converted for use on the 1906A. The main difficulties are its use of COMPLEX*16 variables and other non-standard features and the usual difficulties with short integers on the 1906A. Work is also needed to extend its capabilities from ANSI FORTRAN to the 1906A Extended specification.
XRAY overlay experiment (RET). The XRAY system has been re-implemented without overlays. A number of problems, particularly connected with exofiles, had to be resolved before a successful implementation was achieved. As a by-product, an interesting comparison was made between an overlaid and non-overlaid version of the same program. The figures were:
Overlay | Non-overlay | ||||
---|---|---|---|---|---|
JOBTlME | 17 secs | 16 secs | |||
Core used | 46K | 65K | Clocked time | 9 secs | 17 secs (!) |
MAXQUOTA | 36K | 37K | |||
PAGETURNs | 179 | 327 |
Deductions have yet to be made on these figures.
The manual returned from the printers during this quarter and has been issued. Some new users have begun to make use of the system and this will no doubt highlight any outstanding bugs. One difficulty already encountered concerns the naming of routines, since a user had inadvertently provided a duplicate of one of the system routines (a list is provided in the manual). It is debatable whether a system such as SPROGS should have routines with meaningful names (easier to remember and debug, but prone to duplication) or have standard, meaningless names.
Work is in progress to produce a set of high level routines to perform some of the functions required in animation. To date, the following 7 have been written:
A WIPE is currently being produced. Each of the routines above adds a SPROGS picture file to the sequence list which allows any functions to be used in parallel with any other process.
The picture files change the limits of the region in which the function is taking place and eventually they will be kept in one of the SPROGS file libraries in the same way as the fonts.
Another SPROGS display routine has been written to produce fairing values dependent on the total number of frames for the motion and the present frame number during the motion.
Following the introduction of irregular regions (see previous QPR), a routine BNDRY has been produced to draw the boundary of a region specified by a NULL file. It is also possible to read data from such a file using the NULVAL function, in addition to accessing it by the index variable triple in AR (the AR function can appear as an argument to a display routine and allows data to be read from a NULL file at run-time).
AHF is taking charge of the development of further facilities on the 1906A.
BSI (JRG, RET). During this quarter, the BSI worked long enough to try a simple PDPI5-SPROGS connection. Although the test was somewhat contrived, it did show the possibilities. Data from the PDP15 input devices was written to a communications file in the 1906A. This was read by a SPROGS program, which generated a Tektronix file. At the same time, the Tektronix was logged into MOP and a LISTFILE issued. The resultant picture appeared on the Tektronix screen as soon as the file was released. The Tektronix used was beside the PDP15 (full circle!). It is hoped that a more serious test will be possible shortly.
Further hardware problems are apparent. Currently, an intermittent fault causes the PDP15 to attempt to send to the 1906A without the 1906A program being notified of the attempt. A software time-out has been included in the PDP15 software to re-try transfers which masks the problem. There are also problems when the VT15 display is used at the same time.
Documentation for the PDP15-1906A system has been produced (SPROGS Technical Paper 21). At, present, the system is mainly used for listing large PDP15 files on the1906A lineprinter.
Contrary to statements in previous QPRs, the next link software will be a 2-way character system to be controlled by a PDP15 user under DOS. The advantage of using DOS is that the existing programs and systems under DOS, including PIGS, can be used to carry out some realistic trials of the link software. The hope is to complete the system in the next quarter.
Operator training on PDP15 (JRG, WDS). A fourth operator, Tony Burraston, is gaining experience of the PDP15, as recent shift changes have left one shift without a PDP15 expert. As a programming project, he is writing some subroutines to control the Tektronix storage tube that was connected to the PDP15 last quarter. Using these routines it is possible to display lines, display text, determine the alpha cursor position, determine the crosshairs position when the user strikes a key and continuously determine the crosshairs position under program control, thus providing a crude tracking capability.
One consequence of having a basic routine available to send any character code to the Tektronix from the PDP15, is that one can discover what the Tektronix does in certain situations where the manual is unclear. This has proved useful when writing routines to control the Tektronix from MOP.
Other operators' projects have not unexpectedly made less progress since the operators concerned have returned to their shifts.
PDP15-RSX (RET JRG). A supported version of RSX PLUS III is still not available from DEC. Our current version is still the pre-release received in January which has the bugs and serious problems outlined in the last QPR (Section 6.4).
The bugs are not being corrected by DEC until the supported version is available. In the meantime RSX is only used for system development with other work continuing under DOS.
In spite of the deficiencies of the pre-release software (and the half-hour period it takes to switch from the old DOS to the new RSX system), it has been possible to provide new tasks. These are all related to the needs of the development terminal system (TDV). It is now possible to list device assignments on this terminal and reassign a subset of the device numbers (LUNs) from TDV. A modified version of the MCR task OPEN (which allows alteration of any core or disc location), called ALTER, allows modification of a specified task, making allowance for NORMAL mode and BANK mode addressing.
Work is in progress to implement a user modified version of INSTALL which provides automatic saving of the system. At the moment, saving is a problem (see previous QPR).
In order to assist the Graphics Section in their forward planning, a one-day seminar will be held at ACL in October. A survey of current systems and hardware, together with talks by various users will be given, and discussion sessions will be organised to allow our existing users to indicate where they would like us to put our major effort. At the moment, future plans involve the enhancing of user facilities in SPROGS, and the linking of the PDPI5 to the 1906A. Paul Nelson has agreed to act as chairman, and the initial response from users has been encouraging.
Towards the beginning of this quarter a selection of some twenty ideas for a logo sequence were shown at a BSG meeting, and many suggestions were made by members of the group. It was decided to experiment with shading-in parts of the logo and, even though this proved to be extremely expensive as regards 1906A time, a film of seven examples was sent to Swift Film Productions for colour processing. A final sequence should be selected in the near future.
Following the success of the previous instant film, Professor Edwards has suggested a further 4 films. These will show various different initial conditions, including a chain with a knot in it, and also a different type of chain. This last has two distinct types of molecule, each of which can apparently occupy the same position as others of its own kind, but not of the other kind. Progress is being made on these, including an introduction and a final results table. Difficulties have been encountered with the speed of motion of the chain, and with Professor Edwards' colleague changing jobs. It has also been possible to enhance the final table, following a suggestion from the colleague, by having it flattened by a Monty Python foot!
Some rewiring has been done to correct a few minor hardware problems. Some trouble was experienced with a small regulated power supply, but this has now been overcome. The digital to analogue converters (DAC) have been calibrated to give ± 2 volts, centred about 0 volts.
Experiments have been performed to discover the relationship between the outputs of the DACs and the frequency of the oscillators. The relationship was found to be logarithmic, which is convenient since musical notation is also logarithmic. This enables calibration programs to be written and it is now possible to output any frequency in the audio range fairly accurately.
Software has now been written to allow simple tunes to be played by specifying the start time, duration. and pitch of each note. This information can be read from a disc file, or duration and pitch can be input from the teletype keyboard. Work is almost complete on the first film soundtrack. This is for part of FRAH's Hash Table film. The data describing the notes required was output from the GROATS program on the 1906A which made the film and transferred to a PDP15 disc file. This data was then used to drive the synthesizer and tried with many different parameter settings on the synthesizer.
A FORTRAN subroutine to read an SD4020 tape has been written. It can be used as part of SPROGS or independently. Each call that is made returns to the FORTRAN program a code (line, symbol or text etc) and data concerning the next item on the tape.
If appropriate programs are written, this subroutine can be used to translate SD4020 pictures to other devices or to read an SD4020 picture into a picture file.
Documentation is being produced and the routine will shortly be available in the SPROGS system.
Operations Branch are now mainly responsible for the introduction of new lines and maintenance of the existing communications system.
Considerable rewiring of the communications switchboard has been undertaken by NMP and Rutherford Laboratory. The new bank of plugs and sockets has been installed for the proposed telegraph scanner and to accommodate the new offices. Again, this was built by RL.
The final version of the 7020 emulator was released in mid-June to Southampton, RGO and Sussex. This version has been working satisfactorily at the sites and no faults have been reported. The version includes the following enhancements:
A line to the 360/195 has been used with a pair of Racal Milgo base band modems to allow the connection of the 2050 to the 360/195. The system was made to operate in early June. The operation of the system is being. examined. in connection with the Front End Project. Observations confirm that much better line utilisation is obtained with HASP interleaving protocol. The console response is very rapid whereas under GEORGE response is very slow - 1 sec as opposed to 20 secs. The continued operation of the peripheral during operator interaction is very noticeable.
To gain further information, the code is being modified to store the line traffic and dump it for analysis. 6K bytes are used for this which is more than adequate. It was found difficult to modify the print routine to produce information during operation of the link. This was achieved by examining the register dump area and overwriting it, which prevented use of the lineprinter by the RJE software but allowed its use for dumping. Restoring registers allows the RJE software to continue as if nothing had happened.
This will be delivered in early July. Effort is being divided into four areas:
The 4080-1906A interface is almost complete. All the boards have been delivered and bench-tested. The backplane is almost complete. Testing of the assembled interface should soon commence. The only outstanding item is the delivery of a 100-ft interface cable to the 1906A. This is not critical yet.
Discussions with Rutherford Laboratory have continued and agreement has been reached on required changes to HASP to allow the multiplexing of RJE stations onto one line. Similar changes were originally implemented on the ERCC 360/155.
Discussions with RL, ERCC and others on networks have proved very valuable and continued interaction is envisaged.
Implementation of a Management Game, (Software - Practice and Experience, Vol 4, No 2), R E Thomas and D C Toll