Report by Arthur Payman of a visit to Cambridge by at least him and engineer Peter Titman, on 9 June 1958 to see EDSAC II. On the visit they were given a tour of the machine by Eric Mutch, followed by a technical discussion with Bill Renwick.

Transcript :

From: Mr. Payman. Leo Programming, Minerva Road. 17.6.58.
To: Mr. Thompson, Mr. Barnes, Mr. Pinkerton, Mr. Masters, Senior Programmers.
I attach a report on a visit made to EDSAC II at Cambridge on Monday, 9th June. Mr. Titman was present to handle the technical side of the visit.

[Page 1]

Visit to EDSAC II. 9th June, 1958

This visit occupied a whole day. In the morning a general tour was made of the machine under the guidance of Mr. Hutch. The afternoon was spent in a more technical discussion with Mr. Renwick. The following points are raised as being of particularly programming interest.

1. Edsac II is a parallel machine with a core store which has a very extensive order code and yet contains no more than thirteen different units. This simplicity is achieved by a technique of microprogramming using a core matric. Every action in the code is microprogrammed, i.e. no action is regarded as simple and given autonomy. The steps of a microprogramme (the ‘micro-orders’)  are, therefore, extremely simple, a step in general being a transfer of information from one register to another. The term ‘register’ here is used in the LEO sense, but includes not only the few registers to which the programmer has access, nut also all the registers of the arithmetic unit and coordinator which are technically necessary to the operation of the machine. Thus, for example, the action ‘replace’ (clear, add) if considered starting from the point where the operand is already in the transit register, requires a micro-programme of several seps to control the movement of the operand through the various registers of the arithmetic unit.

It is, therefore, misleading to think of a microprogramme as a built-in sub-routine, or of a micro-order as being basically similar to a macro-order, but, perhaps, slightly simpler. The function of a micro-order is most usually to operate a gate. A set of micro-orders may control a sequence of such operations such that the gross result is the effecting of a macro-order.

The Synthesis of a complex action out of more simple actions is not micro-programming, though such an exercise may have to precede the construction of the micro-programme.

In Edsac II virtually all machine operations are micro-programmed. The passage of information through 'control' determined by micro-programme, determines the co-ordinator cycle.

The Edsac II micro-order is currently timed at 2 1/2 u.secs. It is intended to last 1 1/2 u.secs, and work is in progress to achieve this.

2. Edsac II is equipped with a 1000 characters/sec paper tape reader, designed at Cambridge. This is to be put on the market by Elliotts. Some delay is being caused by re-engineering, since the prototype is set for 5-hole tape only. The production model is to be immediately variable to use 5, 6, 7 or 8 hole tape.

[Page 2]

The reader has been found very satisfactory in use. However, if a very large tape is fed on a job with a particularly awkward rhythm, breaks are liable to occur. This has not yet proved sufficiently troublesome to make the addition of servos worth while.

The reader seems to be a very neat and well designed job; loading is simple, and tape visibility could hardly be improved.

3. Operating controls and monitor facilities have been very much simplified on EDSAC II . Basically, the only controls are START, RESTART, CLEAR STORE, and SINGLE STEP (LEO terminology). There is also a set of manual setting keys. The only monitor on the control desk is the sequence control register, on neons. It has been found that it is not to be retained in the final form of the control desk. On the other hand all of the register of the Arithmetic Unit and the coordinator, are displayed on neons at the back of the machine, for the use of the engineers. 

Programme fault finding is facilitated by the automatic 'report' facility and by the 'trace' facility. The machine punches a report and halts, on a nonsense order or an over-flow. While the machine is running with the TRACE switch set, a record is kept of the last 48 sequence change orders actually obeyed. This record is kept in the part of the store which contains permanent sub-routines and which is not accessible to the programmer. The record is printed out by a special post-mortem routine.

Machine fault finding is greatly simplified by the design of the machine. Since the equipment for a single digit n (i.e. the equipment for that digit position in all the registers) is concentrated in a single unit, it has been found possible to locate faults very accurately by means of test programmes. In fact mathematicians using the machine are permitted to make unit changes when no engineer is available.

4. Output is by paper tape punch (30 characters/sec), but it is hoped to fit the new 300 characters/sec punch shortly. Magnetic tape is used for internal storage only, the same desk reading and writing. The Magnetic Tape has block markers pre-recorded to a fixed block length, the block length being chosen to suit the particular job.

Provenance :
From David Caminer's papers.