Soon after the merger of the Atlas Computer Laboratory and the Rutherford Laboratory, the Engineering Board visited to see examples of relevant work at the Laboratory, have discussions with senior staff and have a Meeting of the Board the next day. This report is the hand out relating to the work of the Atlas Computer Laboratory staff.
The purpose of this exhibit is to demonstrate the different forms of interaction possible on a large computer. Graphical input and output are of major importance in most Engineering applications. The exhibit will show the range of equipment that can be used as aids to interactive computing.
The simplest form of interaction is to allow the user the possibility of editing his stored programs followed by submission of jobs to a batch queue. This is efficient in the use of computer time as the operating system can schedule the batch jobs to achieve maximum throughput. However, the user must wait for his results from the batch run before any errors can be corrected and the job re-submitted. System facilities, which monitor the progress of jobs and report significant events back to the user's terminal, help to make this a viable and economic method of working.
The alternative approach is to allow the user to edit, compile and run his job fully interactively at the terminal. All errors and results are returned to the terminal as the job is executing. Any errors found can be immediately corrected and the job re-run or re-started at the break point. This provides the user with a significant better method of program debugging but at some additional cost. As the job is run when the user requires it, scheduling of resources is likely to be less efficient.
The exhibit will show the same FORTRAN program being corrected and tested in the two modes of working.
Many jobs require the results to be presented in a permanent graphical form. Usually, this will be done off-line with the graphical output being spooled to disc within the main computer. Output to the pen-plotter or microfilm recorder will then be scheduled depending on such things as paper size and camera availability.
Frequently, the user wishes to sample this graphical output to check that the program has worked correctly before allowing it to be plotted. A convenient method of doing this is to have inexpensive graphical storage tubes attached to the main computer which can be used to display selected plots from the spool. Examples of graphical output in the spool will be shown.
In some applications, graphical input and output need to take place as the program is executing. The next stage of the computation may depend critically on the results achieved so far. It may be necessary to view the output data in a variety of ways before the next step can be taken. Input to the program may be essentially graphical in form. Depending on the size of the calculations involved, it may be possible to run the program on a mini-computer with a refreshed display and lightpen attached. However, if the computation required is significant, the mini will need to be connected to a large computer.
The average user will have special programming problems with interactive graphics. It is necessary to provide him with a library of routines for controlling graphical devices and organising his interaction. These routines can be used as the building blocks in the construction of a particular application program.
This exhibit will demonstrate interactive graphics on a small computer on its own, and also when this computer is linked to a larger one in which the bulk of the computation is performed. Examples of the special software required will be shown.
High quality graphical output will normally be produced off-line by special hardware. The most flexible device currently available is the microfilm recorder which allows output to be produced on film and microfiche as well as paper.
The FR80 recorder will be demonstrated. Its special facilities include the production of microfiche (42X and 48X reduction), colour film using an optical filter system, and grey-level output using its 256 different intensity levels.
Examples of recorder output will be shown, including animated film sequences demonstrating the work of a number of university users in engineering applications.
Two methods of developing and running FORTRAN code on the ICL 1906A are demonstrated. In a short demonstration only a small subset of the available features can be shown. Moreover the program being developed is somewhat contrived to guarantee the performance, none the less the distinction between the two approaches should be clear.
In this mode of working, the user prepares and alters his program at a terminal. When the program is ready for compilation and execution, a command is given to the system to add a job to the batch queue or well of jobs waiting for execution. This command (TASK in the case of the 1906A) allows a rich variety of parameters to control the job compilation and execution when it begins. The TASK system has been developed at the Laboratory and is aimed at providing the facilities required by the sophisticated user whilst being easy to use.
A job may wait in the batch queue for several minutes or hours - depending on the machine loading and the job's characteristics. Meanwhile the terminal can be used for other purposes. When the job has finished, the results can be listed on the terminal. (If the user is still logged the system will tell him as each of his jobs finishes.)
It is possible to interact with the resultant programs or even the compiler but the terminal will merely imitate a card reader or lineprinter and it is not easy to obtain meaningful diagnostics.
The advantages of such a system are that the computer can schedule its resources with great efficiency. A large number of users can get an adequate service.
Other documents available from the Atlas Laboratory:
Here the user interacts with the FORTRAN compiler as the program is developed. Program development can be much faster with this mode of working, although it is much more expensive in computer time and time on the terminal. The system is carefully designed to assist the user should any errors occur at any time.
The user logs in as before. This time the compiler in loaded into store (if not already there) and a dialogue takes place between the user and the compiler.
In this interactive system the program so far developed is held in a file in a compiled form and there is no need to recompile the code at each session. Once the program is brought into store the user can either modify the code, add to it or execute it.
During the input of code it is checked as far as possible and the user is given the opportunity to correct lines with faults.
During execution any errors encountered cause execution to halt whence the code can be modified or variables altered and execution can continue.
There are a dozen or so commands which control various means of tracing the progress of code and controlling the system in various ways.
Further details of the 1906A interactive compiler can be found in the ICL publication TP 4275.
The program used in the batch and interactive demonstrations was written to solve a three dimensional Magneto Hydrodynamics problem using the model of Coppi, Greene and Johnson (Nuclear Fusion 1966). The program is about 1000 statements long and was developed by Jean Crow of the Atlas Laboratory.
Since the interactive compiler has to cope with the insertion of code possibly after partial execution and also since it deals with very sophisticated fault finding systems it is slow in compilation and the resultant code is also slow in execution. The complete program takes about 20 seconds to compile but in the demonstration only a few seconds are used as only the code typed in is compiled and the majority is read from a file in compiled form.
The batch compiler is very fast in compilation and execution. The compilation time is 5 seconds. This could be reduced substantially in practice since one would expect well tested parts of the program to be in a library in compiled form.
The interactive FORTRAN would not be suitable for large scale problems although it could help substantially with their development. For this type of work on the 1906A the batch FORTRAN would be used but it is easy to achieve interactive input/output with this system but it is not possible to alter code once compiled. This will be demonstrated in another exhibit.
The program MOLECULE allows one to interactively manipulate displays of molecules. All parameters concerned with the display of the molecules are under the control of the operator.
Commands to the program are typed in by the operator on the PDP15 computer and passed to the 1906A computer for processing. Instructions for drawing the current display of the molecule are sent from the 1906A back to the PDP15 computer, which interprets these instructions and displays the current view of the molecule on the VT04 display screen. Commands issued by the PDP15 consist of an 8-character command name followed by parameters associated with that command.
Commands may be grouped into nine categories.
The 1906A requires that a command be issued by the operator to assign the file containing the molecule description. This molecule description can then be loaded into core store by the 1906A program. Other commands in this category include ones to write out intermediate coordinates and input secondary molecules (more, than one molecule may be displayed on the screen simultaneously; these other molecules are termed secondary ones).
Control commands are concerned with which portion(s) of a molecule is to be displayed. For example, in the case of protein molecules it is possible either to display only those bonds which make up the backbone of the molecule, or to display only those bonds in sidechains. Alternatively, both sets of bonds may be displayed simultaneously.
The operator has full control over the display generated for the VT04. For example, the scale of the molecule nay be changed continuously. Depth-cueing can be used, in which the intensity of each bond displayed is set proportional to some function of the Z coordinate of the midpoint of the bond. This type of display removes the ambiguity in the direction of rotation normally found when one views a rotating stick representation of a molecule. Commands to alter the number of parallel lines representing each bond and the separation between them may also be issued.
The molecule may be rotated about any of the three orthogonal axes, either by a set amount, to position it to a given point, or continuously, to give a three-dimension illusion. It may also be oscillated about a fixed point, again to give an illusion of three-dimensions but retaining some important feature of the molecule at the front.
The conformation of the molecule may be changed at will; one can rotate about any of the single bonds in the molecule, either to a set position or continuously. This facility allows one to interactively explore different conformations of the molecule as one is able to with a mechanical model.
The molecule may be translated by any set amount to position it on the VT04 display screen. This facility is normally only used when one is displaying more than one molecule simultaneously.
More than one molecule may be displayed simultaneously and commands exist to control which of these molecules will be affected by any rotations or translations. For example it is possible to rotate all molecules as one unit or to specify one molecule only to be rotated. It is thus possible to display several different views of the same molecule, starting off with the same input data.
It is possible to label selectively any of the atoms making up the molecule. Labels may be positioned east, south, west or north of the atomic position.
It is possible to edit the molecule description interactively. One can add new coordinates or change old ones if they are incorrect. Bonds may be added or deleted from the list of connections contained in the 1906A.
When one wishes to save a view of a molecule it is possible to generate FR80 orders to reproduce the view of the molecule on any camera associated with the FR80.
High quality graphical output is produced at Atlas using the Information International Incorporated FR80 microfilm recorder which was delivered in the Spring of 1975. This machine is taking over the work previously done by the Stromberg Datagraphix SD4020 recorder which has been in use since 1968.
The FR80 consists of a high-speed computer that accepts and buffers digital graphical information supplied via magnetic tape. This data is processed and displayed as either vectors or alphanumeric characters on a high-precision cathode ray tube. An interchangeable camera system allows the information displayed on the tube to be recorded on a variety of different cameras.
The configuration of the FR80 is as follows:
III15 computer (similar to a PDP 15) with 16K words of 18-bit memory and a ¼ million word fixed-head disc. Data is input via either a 7-track (556 and 800 bpi) or 9-track (1600 bpi) magnetic tape deck.
A Precision Light Source cathode ray tube with white (P24) phosphor for colour recording. The standard FR80 contains both a vector and character generator. The Atlas machine also includes the following options:
The plotting area of the FR80 is divided into a raster of 16384 points in each direction. Vectors can be drawn between any two raster positions on the plotting area at any one of eight line widths. Four different vector drawing rates are available with the slowest giving the greatest precision. It is possible to resolve 72 Line pairs per millimetre on 35mm film.
Characters can be displayed at any raster position. Instead of a hard-wired character set, the character forms are stored in the III15 core store in a compact form. Up to 256 characters are available at any one time and these can be displayed in 64 different character sizes. A number of different character fonts are available. It is also possible for the user to define his own.
The following cameras are available:
A III 5010 film processor was purchased with the FR80. This can produce good quality negative or full reversal processed film (16, 35 and 105mm). The speed of processing is about 15 feet per minute.
The film entitled Finite Elements was made jointly by ACL and the Royal College of Art, using the ANTICS Computer Animation package on the ICL 1906A and the SD4020 plotter. This 10 minute colour film was designed for two main types of audience:
and serves also to demonstrate the suitability of computer animation techniques for the presentation of technical information.
The first part of the film (not being shown in this brief demonstration) gives an introduction to the method itself, the equations to be solved, and the matrix techniques commonly used. A variety of different types of element - rods, triangles, wedges, are described, and the necessity of achieving a balance between high accuracy and fast execution of the computer program is discussed.
The second part of the film examines two application areas in more detail.
This problem was solved by Nottingham University's Mechanical Engineering Department using their FE package PAFFC, and their division of the model bridge into finite elements and the resulting stress contours are portrayed. The need for a very large computer to tackle this kind of problem is emphasised here.
Two problems solved by Southampton University's Civil Engineering Department are shown - firstly a case of fluid flow round an obstruction in a channel, in which finite element analysis is used to compute streamlines and other flow variables, and secondly a study of the dispersion of pollutants in the Solent given the location of the pollutant source, tidal data etc.
It is hoped that, although a short film, Finite Elements will satisfy the objectives of its makers in that it will show engineers and those concerned with engineering computing the versatility of the method and the comparative ease with which it can be used to further engineering research into many present-day practical problems.
Engineering Board Members | ||
---|---|---|
1. | * Mr J M Ferguson | Chairman. Engineering Consultant |
2. | Professor G Allen | Professor of Chemical Physics, Manchester University |
3. | Professor R S Benson | Professor of Mechanical Engineering, UMIST |
4. | Professor J Brown | Head of Electrical Engineering, Imperial College |
5. | Mr F D Boardman | Engineer, Central Electricity Research Laboratory |
6. | *Dr J H Horlick | Vice-Chancellor, Salford University |
7. | *Dr A J Kennedy | Director, BNF Metal Technology Centre |
8. | *Dr D J Lyons | Director General of Research, Department Of the Environment |
9. | *Sir Ieuan Maddock | Chief Scientist, Department Of Industry |
10. | Dr N G Meadows | Deputy Director, Glasgow College of Technology |
11. | Mrs M K McQuillan | Research Manager, Imperial Metals Industries Ltd |
12. | Professor B Randell | Professor of Computing Science, Newcastle University |
13. | Professor H H Rosenbrock | Professor of Control Engineering, UMIST |
14. | Mr A G Senior | W S Atkins, Research & Development |
15. | Mr J A Stokes | Engineering Consultant |
16. | Professor W L Wilkinson | Professor of Chemical Engineering, Bradford University |
17. | Sir Richard Young | Former Chairman of Alfred Herbert Ltd |
Assessors | ||
18. | Mr A Callaghan | University Grants Committee |
19. | Dr R L Lickley | Director, Hawker Siddeley Aviation |
20. | Mr P H Stephenson | DSCO, Department of Industry |
Secretary | ||
Mr H J W Shepherd | SRC, London Office |
* indicates Member of SRC Council
For the visit, Rutherford Administration attached the Who's Who entry for each member. These have not survived the ravages of time too well. Some information from Wikipedia etc is given below.