I started work, but went straight back to school. The learning curve was extremely steep.
ICL had been formed in 1968 by the forced merger of ICT and English-Electric-Leo-Marconi. ICT had itself been a merger between Powers-Samas and British Tabulating Machines (a spin-off from IBM), English-Electric etc. was clearly another merger. Leo, you may remember made the first commercial computers for the Lyons Company, a chain of Tea Shops. The British computer industry had been a mess.
Up stepped Tony Benn (formerly Viscount Stansgate), then Minister for Trade and Industry in Harold Wilson’s Labour government, and, forsaking the obvious money-sink that nationalisation would have meant, forced the industry to consolidate. With 20/20 hindsight, he made the right decision at the time.
ICL made computers, I guess there may have been people who called it an IT company, but we didn’t. We made computers. In fact, for a small player on the world stage we made lots of different computers, and all the software for them. From ICT we inherited the 1900 series, with a 24bit word and 6-bit characters, and from English-Electric we inherited the System 4, an IBM 360 clone.
The 1900 series had sold well to the UK government sector, and it was to protect supplies of this range that I imagine Mr Benn took his decision. Clearly though rationalisation of the ranges was a priority to cut costs and improve marketing effectiveness.
Enter the New Range, as it was called. New Range was to be a proper 32-bit architecture, designed from the ground up to be something that could evolve to meet any computing needs. We proposed a range of machines from small ones with, say 50 kilobytes of main store, to 1 Million-instruction per second super computers with a massive 1 or 2 megabytes of store. You have to remember that memories were made in those days by threading 2 mm ferrite toroids onto wires like a kind of ferrite carpet, and every 4 (that’s 4 , not 400, or 4 million) and-gates in the cpu were an integrated circuit. Cpus in the Megahertz-range needed massive water cooling systems and occupied some of the largest air-conditioned rooms on the planet.
After company indoctrination, I was put to work as a test programmer on New Range. I would work on the baby computer end of the range.
The real fun testing a new mainframe, with a new architecture is that there is nothing to start from. The machine didn’t even have a bootstrap. To load a test program, it would be cross-compiled on a 1900 and punched onto 8-track paper tape. Physically, (we didn’t have networks then,) you take the paper tape to the new machine and put it into the reader. Now you use the hand-switches to enter a simple bootstrap loader (in binary) into the memory, which takes a few minutes, and is very error prone. All the bootstrap does is to read the papertape into memory and enter it. The first tape loaded is the full bootstrap program, which you then use to load your new test program. Oh, and the user interface? 32 or so switches and 32 or so lamps. You learnt to play it like a piano.
By the time the first year had gone by, the baby new range machine had been cancelled, and I’d moved onto the next machine in the range. At about that time too the company decided the New Range was going to take longer than they thought to get market ready – the Operating System was the big issue. So as a stop-gap a new 1900 ordercode machine was launched – this was to become the 2903, my first introduction to a microcode engine.
There are a number of ways that a cpu can be made to obey a set of machine-code instructions. You can design the hardware to do it natively – the instruction decoder logic breaks down the machine code bits as switches to drive the adders and other parts of the mill – or you can design the hardware to obey a limited machine code, and use that to write an interpreter for the target order code. Many mainframes since the early 1970’s have been of the latter type, and recently many microprocessors are too. In general the native order-code that the hardware executes is some form of limited instruction set code.
The 2903 used a mill called ‘Micos’ which had been designed as a research idea into fast microcoded cpus. Don’t confuse the words ‘Microprocessor’ and ‘Microcode’ – as we shall see this led to some interesting times for me later on…
I worked a little on some tests for 2903 and it was duly released and achieved a lot of success in the market. It brought the cost of a 1900 system in range of medium-sized enterprises.
The New Range developments continued and we were beginning to work on the peripheral controllers. Today we think of peripheral controllers as single-chips that occupy maybe four square centimeters of real estate. In the 1970’s, peripheral controllers were large cabinets of complex electronics built from many printed-circuit boards. A single peripheral controller for New range was as big as my desk.
ICL planned to use their existing small computer designs as the basis of the New Range peripheral controllers. The Micos engine used in 2903 was to be used as the controller for communications links, the CPC (Communication Peripheral Controller). With my (mininal) experience with 2903 micro-code I was put to work testing the comms links on a CPC. Suddenly I was an expert in comms protocols. (Well can you remember what HDLC protocol was for?)
By this time too, I had been promoted to Team Leader, and had a one or two new graduates under my wing. I’m not sure who learnt most from whom – in the three years since leaving University the whole idea of Computer Science as a course had been accepted and these people knew more theoretical computing than I did. Still turning theory into engineering has always been difficult – we managed.
However, all was not well with ICL, even with the 60% gross margins on computers enjoyed back then, there was not enough coming in to cover the massive R&D costs. In 1975 the Stevenage Labs were closed and all computer development moved north to Kidsgrove and Manchester.
In 1974 I had married Ros Arlow, a girlfriend from University, and we were living in Stevenage in a small flat. We had plans to buy a house, but the cash situation was not good. The alternatives for me in the reorganisation were either to take voluntary redundancy, or to relocate at the company’s expense to Kidsgrove.
Now I’ve been to Kidsgrove many many times, both before the event, and since, and although it’s a pleasant enough place, it holds little charms for me, and Ros hated it. We decided that redundancy, with it’s cash handout, offered the opportunity to buy a house, so that’s the option I took. I got a new job in two days, left ICL on a Friday with a few thousand pounds, and started work the next monday – I got a pay rise too.