DEUCE stands - or stood - for "Digital Electronic Universal Computing Engine" - quite a neat acronym, and better than many current ones.
I worked on a DEUCE for about two years, from 1963 to 1965, and it was my first exposure to a digital computer. Back in those days, when I studied mathematics at university, computers were thought undignified by mathematicians, and were the responsibility of the "dirty paws" in the engineering department.
DEUCE was characterized by two levels of storage, much as in modern machines. The high speed storage was a set of mercury delay lines of varying sizes. These are tubes containing mercury, into which a stream of sonic pulses representing each bit is injected by a transducer at one end. A microphone at the other end received the pulses a few milliseconds later, and they were then processed and modified if needed, squared up, and fed back into the transmitting transducer. Data could thus be recycled indefinitely. These delay lines were the forerunner of core memory, which had not yet appeared. As I remember, the delay lines held various multiples of 32 bits, and each machine had about half a dozen of them.
Nowadays, people would ask awkward questions about the level of mercury vapour in the machine room. Back then, no one seemed to worry about it.
The second storage level was more similar to modern devices - a magnetic drum not unlike a present-day disk. Capacity was less however - I remember it as 48K 32-bit words - about 200 kilobytes in modern terms.
The other devices were a card reader and a punch. No magnetic tapes - all programs were loaded stand-alone from cards.
The card format was also interesting. They were physically identical to the standard IBM 80-column punched card, and the IBM scheme for coding data and text could be used. But binary programs were encoded onto the cards using all the positions, achieving a good level of packing, but making the cards almost impossible toproduce or modify by hand. This did not stop people from trying - I remember watching others using a handpunch to insert one or two extra holes in carefully-chosen places, and even filling in unwanted holes with debris from the hand punch, and a little saliva to fix them in place. Reliability of this procedure was well known to be poor.
The binary instruction format was very complex. Each instruction occupied a single row on a punch card. Besides specifying an operation code and the operands, (typically the identity of the delay line from which data was to be taken or where it was to be placed), there were also a 'wait number' and a 'timing number'. The wait number specified the number of clock cycles before the required data would emerge from the delay line, and the timing number was the number of clock cycles to wait before the next instruction would emerge from the delay line containing the executable code. Consequently, optimising a progam meant arranging the instructions in the order that would minimize the clock cycles that were wasted in waiting. This meant in turn that the instructions would not appear on the punch cards in the order in which they would be executed, adding a great deal of complexity form a human point of view.
The machine did not have an operating system in the modern sense. There were two primitive languages; Alphacode and Easicode. I think the first was interpreted, and the second compiled. They both simulated a set of 32 bit registers on which arithmetic operations could be carried out; all statements looked like
A1 = A2 + A3Writing this, I find my memory is much hazier than I thought - it was well over thirty years ago after all. If anyone has more precise details, I would be delighted to hear them, and would correct and expand this description.
In April 2003, John Barrett also contacted me from Australia, and told me about his much more detailed web site about the DEUCE. The site can be seen at http://www.users.tpg.com.au/eedeuce