Although UTECOM was officially opened on 11 September 1956 by the Premier of New South Wales, Mr Cahill (NSWUT, 1956), the magnetic drum was not then operational (NSWUT, 1957). The installation of UTECOM at the Kensington campus of The New South Wales University of Technology took place during August, September and October of that year, and it would appear from the first annual report that it was fully commissioned and had passed its acceptance trials by 16 October 1956 (ibid). UTECOM was upgraded to a Mark I DEUCE around 1958 with the installation of the Automatic Instruction Modifier (AIM) and an enhancement of the card input/output system to read and punch 64 columns instead of only 32. In 1963 the magnetic drum storage was enhanced, and a teleprinter incorporating 5-channel paper tape input/output was attached.
UTECOM was procured for the purposes of research and teaching. Acquisition was made possible through a state government grant.
There is some ambiguity as to the relative operational dates of UTECOM and Sydney UniversityÕs computer, SILLIAC. UTECOM was officially opened on 11 September 1956 by the Premier of New South Wales (NSWUT, 1956), SILLIAC on 12 September 1956 by the Governor of New South Wales (Pearcey, 1988). (de Ferranti, 1956, p. 9) stated that the first SILLIAC program was run on 4 July 1956. (Pearcey, 1988, p. 32) nominated 24 June when the first SILLIAC program was run, while (Bennett, 1963, p. 13) nominated July. (Millar, 1987) states that the first program was run on 4-5 July (p. 28), and that it was operating on open day, 22 July (p. 30). Millar also states (p. 28) that "as soon as the machine was made operational . . . this was the first program carried out, on 4-5 July 1956". It would appear that the machine had yet to undergo proving trials. The implication is that the principal part of the machine was not operational until quite some time later than either of these two latter dates. It is worth mentioning that SILLIAC was fabricated on site in the Physics building at Sydney University, whereas UTECOM was built in England at the Kidsgrove factory of English Electric. Thus, UTECOM was entirely functional and had passed its acceptance trials at Kidsgrove prior to being shipped well before the September date, and in all likelihood, even before the July date ascribed to SILLIAC. It should be pointed out that SILLIAC contained 1024 words of memory. As initially installed, it did not have a backing store (a magnetic tape backing store was not installed until 1959 (Millar, 1985, p. 60)). The DEUCE, on the other hand, was provided with magnetic drum backup storage as standard equipment. Nevertheless, in this account, SILLIAC has been accorded the status of being the second operational machine. However, UTECOM was the second machine containing a backing store to be installed in Australia.
In this monograph are provided brief technical descriptions of the DEUCE architecture, hardware, computational units including Control, programming, an outline of the main programming aids used and/or developed at the UTECOM Laboratory, including assemblers, interpreters, control programs and compilers. Remarks on advanced features are included. Operation, maintenance and staff are briefly discussed.
The illustration is of the DEUCE console.
The DEUCE was a general-purpose vacuum-tube digital computer, with a serial organization and a 1Mhz clock rate. It was a re-designed Pilot ACE, with a different combination of registers and renumbered addresses, and with significant differences in the hardware implementation and construction intended to increase reliability and to render maintenance easier. (The 1Mhz clock rate was envisaged by Turing for the ACE, and was already proven in the Pilot ACE). The Pilot ACE and DEUCE were so similar that Pilot ACE machine-language programs could be converted to DEUCE programs by the routine substitution of new register addresses. For an outline of the rationale for building DEUCE, see (Lavington, 1980).
The word size was 32 bits, and the machine's arithmetic units were capable of performing single, double, and mixed-precision binary integer arithmetic. Negative values were held in two's complement form. A hardware unsigned integer multiplier and a signed integer divider were included.
The DEUCE had 18 registers, a fast main store of 384 words (256 words of execution store, 128 words of auxiliary store), and magnetic drum storage of 8192 words. The hierarchy of registers comprised four 32-bit registers including an accumulator, three 64-bit registers including a double-precision accumulator that could also be used as two single-precision accumulators, and two quadruple registers, both holding four 32-bit values.
The fast main store -- "a large store for a machine of its class" (EECo, 1958) ¾ was organized as 12 banks or delay lines of thirty-two 32-bit words each. Eight banks held executable instructions, and 4 banks comprised auxiliary storage. One of the delay lines could function as a push-down stack and as a first-in first-out stack.
The magnetic drum storage unit, or, to give it its correct name, the Magnetic Recording Drum, contained one block of 16 reading heads, and another separate one of 16 writing heads. Each head provided access to a track of 32 words, and both blocks of heads could be moved independently to any of 16 positions. Thus the heads provided coverage of 256 tracks, each containing 32 words. An entire track of 32 words had to be transferred during a magnetics read or write operation.
Because the magnetic drum was synchronous with the system clock, data transfers commenced immediately, and, at the rate of 9ms per bit, a track of 32 words could be transferred in 13ms (including safety margins). The time allowed for positioning the read and write heads was 35ms (though the actual time was approximately 22ms). Each block of read heads and write heads could be positioned independently, but only one track of data could be read or written at any one time, as there was only one main-store buffer. An electronic interlock prevented the user from accessing the magnetic drum buffer, or from moving the heads, until the current read or write operation was complete. After the installation of the site-built (but standard) rationalized magnetics unit in 1963, the read and write heads could be moved simultaneously, the write heads could be moved while reading a track, and the read heads could be moved while writing a track, and other unnecessary magnetics interlocks previously imposed on accessing the drum buffer were removed. These modifications increased the speed of programs that used the magnetic drum intensively. A unit to check magnetic drum read operations and using a double strobe technique (achieving the effect of a double read) was installed in 1963. The DEUCE magnetic drum storage device -- unique on account of its moving heads -- was the forerunner of the moving-head disc. Measuring 4" (10cm) in diameter and 6" (15cm) in height, the drum rotated at 6510 rpm. The head gap was 0.001" (0.025mm).
Within a few months of the first DEUCEs going into service, it was realized that the machine would benefit from an auto-increment and decrement facility (the Automatic Instruction Modifier, or AIM) associated with the quadruple registers (Burnett-Hall & Samet, 1959, p. 43). Without the AIM, instruction modification -- a necessity in the DEUCE architecture -- tied up either or both of the single and double accumulators. The AIM was needed to deal with 64-column input/output (outlined below), if for no other reason, because two accumulators would otherwise have been needed to deal with reading and punching binary words. The AIM and enhanced input/output were installed on UTECOM in about 1958.
The peripherals were Hollerith 80-column punched card equipment, the reader speed being 200 cards per minute, and the punch speed being 100 cards per minute. Reading and punching transferred one binary word per row of a card, conversion to and from decimal being performed by software. Initially, only 32 columns of each card were read and punched, but in about 1958 a facility to read and to punch 64 columns was installed, doubling the input/output transfer rate. (In this mode, two binary words were transferred per row of a card.) In 1963, a Siemens M100 teleprinter with 5-channel paper tape reading and punching attachments operating at 75 Baud (10 cps), was installed to relieve the load on the card punch, interfaced with a locally designed and built parallel to serial unit.
Time Maximum operation rate ==== ====================== add/subtract/transfer 64µs 15,625 operations/sec block add/subtract/transfer 32 + 32n µs 30,300 operations/sec (for n=32 words) of n words fixed-point multiplication 2080µs 480 operations/sec fixed-point division 2112µs 473 operations/secTimings of the main input/output equipment are shown in the table below:
Component Speed ========= ===== card reader input 200 cards per minute 4,800 binary words per minute 1,400 nine-digit decimal numbers per minute card punch output 100 cards per minute 2,400 binary words per minute 700 nine-digit decimal numbers per minute magnetic drum: rotational speed 6,510 rpm track read time 13ms for 32 words head positioning time 35ms
A full summary of the technical characteristics of UTECOM and its off-line equipment is given in Table 1 in the Appendix.
For a detailed description of DEUCE architecture and programming, see (Burnett-Hall & Samet, 1959) and (EECo, 1956a). The former reference includes a description of the AIM and the 64-column card input/output facilities and programming, whereas the latter reference, being an earlier work, does not.
The DEUCE Mark II had the AIM and an IBM 528 80-column combined reader and punch unit with two modes of reading: standard non-buffered 64-column binary input/output, and a new buffered 80-column alphanumeric to/from 6-bit Binary Coded Decimal input/output. The IBM 528 card reader and card punch could operate locked together (simultaneously) at 100 cpm, or individually at 200 cpm for reading and 100 cpm for punching (but not simultaneously).
The Mark IIA had the AIM, an IBM 528 80-column read and punch unit, and 7 extra delay lines contributing an additional 224 words of main execution storage.
While the Mark I was intended for scientific use, the Mark II was specifically developed for the growing commercial market, although any DEUCE equipped with magnetic tape was suitable for large-scale commercial processing. Nevertheless, the DEUCE Mark II retained all the scientific and commercial facilities of the Mark 0 and Mark I which were upwards-compatible with the Mark II and IIA.
Any DEUCE could be fitted with Decca magnetic tape units, an English Electric paper tape reader capable of reading 5, 7, and 8-channel tape at 850 cps, and a Creed paper tape punch capable of punching 5 or 7 channels at 25 cps (EECo, 1958; EECo, 1961b). The paper tape reader, incidentally, could stop the tape between characters. The Mark II and IIA DEUCEs could be fitted with two IBM 80-column read and punch units. The Decca magnetic tape units were 7-track twin transports, probably of the kind described in (Barron & Swinnerton, 1960), tape speed probably being 100 inches/sec at a recording density of 100 characters/inch.
A DEUCE Mark I cost £52,000 Sterling, and a DEUCE Mark II cost £60,000 Sterling (EECo, 1958). A second magnetic drum, magnetic tape, paper tape, the 7 extra delay lines and a second IBM 528 80-column read and punch unit could be attached as "optional extras". The paper tape reader equipment cost £1,500 Sterling.
In all, approximately 33 DEUCE machines were installed in England, Scotland, Ireland, Europe, not to mention Australia. Users included government, universities and industry. Some organizations installed two and three DEUCE computers. The English Electric Company operated a service center at Kidsgrove with a Mark I and a Mark II DEUCE, and a Mark I at its London Computing Service. The Company had two Mark I machines at Wharton. The Ministry of Aviation Royal Aircraft Establishment installed two Mark I DEUCE machines at Farnborough. The Ministry of Agriculture Fisheries and Food is believed to have operated three. Six machines were used for aircraft design. Educational institutions included Queens University (Belfast), Liverpool University, Glasgow University as well as The New South Wales University of Technology.
The illustration is of the DEUCE console.
A subsidiary bus, the Instruction Highway, connected Delay Lines 1 to 8 to the Instruction Register TS COUNT. Because the data bus could already have been in use for the data transfer specified by the currently-executing instruction, the separate bus, the instruction highway, was needed to send the next instruction into TS COUNT.
Electronic Technology vacuum tube, germanium diode Number of items 1450 approx. Main item types E92CC, E90CC, Z77 (6AM6), 6CH6, ECC91 (6J6), 6060, 2D21 (thyratron), OA81 (diode) Operation mode serial, 1µs/bit Execution Store acoustic delay size 256 words (480 for DEUCE Mark IIA) type mercury main store access time 496ms average; max. 992ms, min. 0ms special 128 words of auxiliary store Registers 4 1-word 3 2-word 2 4-word Backing Store Magnetic size 8,192 words type Synchronous moving-head drum, 256 tracks, 6510 rpm, separate read and write heads access time 13ms read/write 32 words 35ms head shift rotational delay none Word Size 32 bits Instruction size 32 bits Instruction format 2+1 address: 8 segments including Source, Destination, NIS, Director, Characteristic, Wait, Joe, Timing, Go# Number of instructions 1024 approx. Special multiple-transfer instructions, push-down stack, auto-increment/ auto-decrement instructions Speed: 32-bit fixed-point add/subtract 33µs to 64µs 64-bit fixed-point add/subtract 66µs to 96µs fixed-point multiply 2,080µs small integer multiply 96µs to 1066µs fixed-point divide 2,112µs floating-point add/subtract 6ms av. (software) floating-point multiply 5.5ms av. (software) floating-point divide 4.5ms av. (software) convert binary to BCD 2Mc 8mc convert BCD to binary 2Mc 10mc convert binary to characters 2Mc 10mc convert characters to binary 2Mc 10mc convert BCD to £.s.d. 2Mc 10mc convert binary to tons, cwts, qtrs, lbs 2Mc 9mc rearrange bits in word 2Mc 12mc count bits in a word 2Mc 15mc other block move, 33µs per word for 32 words Optimum Coding yes I/O media on-line punched card I/O (80-column cards) teleprinter 5-channel paper tape I/O** input speed 200 cards per minute output speed 100 cards per minute modes: 64-column non-buffered 24 binary words/card 80-column buffered alphanumeric to 6-bit BCD teleprinter, paper tape I/O 10 cps** I/O media off-line card tabulator (printer) 100 cpm, card sorter (600 cpm), card-operated typewriter (10 cps), 2 card reproducers (100 cpm), electric key punch, teleprinter Date Installed Sept. 1956 Comment English Electric adaption of Pilot ACETable 1: Technical characteristics of DEUCE. Sources include (EECo, 1956a; EECo, 1958)
# Director = bits for AIM, NIS = Next Instruction Source, Characteristic = (single, double, or long transfer), Wait = relative time to start the transfer, Joe = loop counter bits, Timing = relative location of the next instruction, Go = instruction to be executed as soon as possible on entering Control.
** The English Electric paper tape reader read 5, 7, and 8-channel tape at 850 cps, and could stop the tape between characters. The standard Creed paper tape punch speed was 25 cps. Programming the UTECOM teleprinter output and tape punch was standard, but programming for the UTECOM paper tape reader was non-standard.
Extant Manuals etc: DEUCE STAC Programming Manual (38 pages); DEUCE General Interpreter Programme, 2nd Edition (DEUCE NEWS 63) (alias G.I.P. 8) (104 pages) 'English Electric' D.E.U.C.E. Programming Manual May 1956 NS-y-16 (70pp) N.S.W. Scheme B Programs (about 10 pages) LR23BM General Decimal (Matrix) Read to Drum (DEUCE Programme No. 738 (18 pages) XR03/1 Multiple Regression Programme (DEUCE Programme No. 730) (11 pages) Royal Aircraft Establishment, "A Programming Handbook for the Computer DEUCE" --D.G. Burnett-Hall & P.A. Samet - 207 pages. English Electric DEUCE Logic Diagrams & Circuit diagrams for Magnetic Tape, some 3 x A4 in size (12pp) Flow diagrams for GEORGE (original manuscript probably c. 1957) (this is the historic translator for Charles Hamblin's zero-address language for a stack-based machine, on which the English Electric KDF9 machine was based -- 74 pages) English Electric General Description of "DEUCE" Digital Electronic Universal Computing Engine (Aug. 1958) (17pp) Various roneod programming manuals produced by the UTECOM Laboratory, for GEORGE, GIP, TIP, SODA (I think) etc. (probably c. 1958) - NSW University of Technology Introduction to High Speed Digital Computing (17pp)George Programming, page 1
- University of NSW Programming Course (57pp) - Alphacode Notes (15pp) - University of NSW High Speed Digital Programming Course (17pp) - The UTECOM General Interpretive Programming Scheme (12pp) - GEORGE and Alphacode Exercises Set II (2pp) - GIP Examples (2pp) UTECOM Laboratory Safety Precautions (4pp) Some flow diagrams for STAC (manuscript form) (about 400 pages) (c.1964) STAC III for DEUCE Mark I and II(A) (12 pp) 31.3.65)(12 pages) R.A. Vowels, UTECOM -- An English Electric DEUCE (1993) (33 pages, including 4 photos)A list of DEUCE programs and Subroutines. This list has been compiled from various sources including the GIP Programming Manual, the STAC Programming Manual, and DEUCE NEWS 18. The list is still in preparation.
Other pages of related interest: John Barrett's DEUCE site
DEUCE MARK II
David Green has compiled an extensive list of surviving DEUCE documents.
Weapons Research, WREDAC, and the Second Australian Computer Conference
Any information on 80-column I/O would be appreciated to allow me to complete the implementation of the Mark II input-output. (see below for contact details)
Updated 20 September 2007.