NUSSE was the Norsk Universell Siffermaskin, Selvstyrt, Elektronisk computer, which was based on an APEXC(N) (sometimes called APE(X)C) in kit form after contact with A.D. Booth at the UK Birkbeck College. A computer at the scale of EDSAC was out of reach financially for Norway at the time, but the APEXC(N) was affordable.
There are two references in UK books by A. D. Booth to this computer,
- "Automatic Digital Computers" 1953 1955 reprint chap 35 p270
- "Up to the present, two copies of APE(x)C are in existence. The first at the Norwegian Board of Computing Machines, Oslo, under the direction of Dr. T. Hysing, is undergoing tests. The second is in general use at the laboratories of the British Tabulating Machines Company."
- "Faster than Thought" 1953, 1955 reprint chap 13 p171
- "A.P.E.N.C. Norwegian Board for Computing Machines
- The machines are properly engineered versions of S.E.C. built with miniature components. The digit repetition rate is 50kc/s. Storage capacity is of the magnetic drum type and gives 512 thirty-two-digit binary numbers. The machine is provided with a high-speed, short-cutting multiplier and takes 500usec for addition and subtraction, and (n x 500) usec for multiplication, where n is the number of "live" digits in the multiplier.
- Only 415 valves are used, including all diode elements. Power consumption is 1.5KW
- Input and output equipment varies with the requirements of the users, three types being available---
- Punched-paper tape read photo-electrically as input, output to reporforator or teleprinter.
- Magnetic-tape input and high-speed output. Printing via a teleprinter.
- Punched-card input and intermediate store. Output via Hollerith tabulator."
Unfortunately the kit turned out to be incomplete, and also not particularly good ("unsound" is the term used in the reference). Thomas Hysing and crew at CCIR (SI) in Oslo got the computer working in the spring of 1953 after a number of modifications and extensions. By 1954 NUSSE was considered sufficiently stable to go to the operational phase, and was moved to the Norwegian Computing Centre (NCC).
- Main memory: Rotating drum with 512 32-bit words (some sources say 36 bits, which could explain the instruction bits issue mentioned below)
- Input/Output: 5-channel paper tape and a Teletype printer. This appears to have been extended somewhat later, see the "Faster than thought" reference above
- 3 main registers: Ar, Mr, Or
The instruction format appears to have been as follows:
- 7 bits for counter (n)
- 10 bits for data address (x-addr)
- 5 bits for operation (op)
- 10 bits for address of next instruction (y-addr)
(The above adds up to 32 bits, there is some unclear issues here (to the wiki article writer) in that the format is also described as having a bit '(b)' which is called Not Used).
As for the address of next instruction, it was common for drum-based computers to have an implied 'goto next instruction' as part of every instruction. It is the drum address of where to fetch the next instruction.
There were 16 basic operations, only one is known at this point: Operation 14 was an operation to transfer the (32-n) least significant bits from register Ar to memory in address x-addr. And it would fetch the next instruction from y-addr, of course.
Basic operation time was 1ms if the next instruction was immediately available under the drum, but went up to 16ms if it was necessary to wait a complete turn of the drum to fetch it. And another 16ms if a complete turn was necessary to access the data address, y-addr. From this it is clear how a dedicated programmer could optimise the program so that the next data and the next instruction were always ready when needed.
The NUSSE was decommisioned in 1961.
- Applications and Technologies for Maritime and Offshore Industries, by Trygve Reenskaug, Department of informatics, University of Oslo