TIP(ID:721/tip001)Tabular Interpretive Autocode for the DEUCETabular Interpretive Program Robinson, English Electric, 1956. An early interpreted language for the English Electric DEUCE, featuring table-oriented operations In many ways this was the first spreadsheet Hardware:
Related languages
References: in [EEC] DEUCE News No. 11 December 1956 English Electric Company view details in [EEC] DEUCE News No. 11 December 1956 English Electric Company view details in The Computer Journal 1(4) January 1959 view details INTRODUCTION The object of automatic programming is, of course, to reduce the effort required to write programs, and this is achieved by using, instead of ihe computer's own code of instructions, a pseudo-cork which is u«"illy closer to the programmer's habitual way of describing the operations which he wants the computer to perform. (The word 'programmer' is used here rather in the sense of'anyone wishing to write a program', than that of 'someone specially trained to write programs'.) A program having been written in a pseudo-code, there are two ways of making it produce the desired result, viz. by interpretation or by translation (or compiling, as it is often called). In the first method, each code-word is interpreted as it is reached in the course of the calculation, and many of them will therefore be interpreted in my times. Since the interpretations occupy computer-time additional 11 11 at required for the actual computation, interpretative schemes produce |> grains which are slower than those written in the machine-code, though in some applications, e.g. in G.I.P. and T.I.P. (Ref. 1), the amount of computation specified by each code-word is usually so great that the interpretation time is negligible. Another feature of pseudo-codes which tends to produce slow programs when the interpretative method is used is that they do not fully reflect the storage-structure of the computer. It is, of course, desirable that a pseudo-code should be as free as possible from this structure, since it is alien to the conventional description of calculations, etc. On the other hand, the structure is designed mainly to achieve as high a speed of operation as possible for a given overall cost of the computer, and the neglect of it must cause a reduction in speed. To satisfy both these requirements, which are incompatible in any interpretative scheme, and to avoid the repeated interpretation of codewords, we can use the computer to translate, once for all, a program written in a pseudo-code free from computer-determined structure into a fast program of computer-instructions, taking the storage-structure fully into account. The problems which are presented in the writing of a program for making such a translation are, however, far from trivial. In many interpretative schemes, some translation is also done, i.e. the original code-words are first translated into others with properties nearer to those of the machine-code, e.g. symbolic addresses may be converted to absolute computer addresses. (We, at English Electric, usually reserve the term 'compiling' for this process, for convenience, although it is not essentially different from translation, as defined in the previous paragraph.) A 'compiler', i.e. a program for doing this preliminary work, is much easier to write than a 'translater', and may save a good deal of interpretation time. Descriptions will be given of four interpretative schemes which have been prepared for DEUCE, one of which has been referred to by Mr. Robinson (Ref. 1). in each case, some compiling is done on the code-words before they are interpreted. The pseudo-code used in the first of these schemes, called George, is unusual, being in fact an extension of a notation ('reverse Polish') suggested for mathematics (Ref. 2). The second scheme, Alphacode, was inspired by the Manchester Autocode (Ref. 3), but attempts to make every code-word a statement in plain English. The third scheme, STEVE, employs a special-purpose pseudo-code, similar in form to those used by G.I.P. and T.I.P., but intended solely for calculations on the thermo-physical properties of steam and water. The fourth scheme, Easicode, is a general-purpose scheme with a form of compiled instruction giving rapid interpretation. There follows a description of an existing translating program whose pseudo-code is known as Soda. Both this and Easicode retain some of the storage structure of DEUCE, which has in some degree facilitated their writing but which makes the writing of programs in these pseudo-codes rather more difficult than in Alphacode, for example. Finally, some of the difficulties which arise in translating from a pseudo-code which is free from computer-determined structure are discussed, together with some of the techniques which are to be used in a program for translating from Alphacode to DEUCE machine-code. in Goodman, Richard (ed) "Annual Review in Automatic Programming "(1) 1960 Pergamon Press, Oxford view details INTRODUCTION Despite the organization of vast libraries of subroutines and programs, and the facilities for testing new programs, a real need has grown up for flexible and powerful schemes, capable of being used to construct new DEUCE programs in a fraction of the normal time. During the last four years a variety of such schemes has grown up for DEUCE, and the present paper and a companion paper by Mr. S. J. M. Denison, reviews some of these, and fits them into perspective. The schemes STAC, G.I.P. (General Interpretive Program), T.I.P. (Tabular Interpretive Program), Alphacode, GEORGE, SODA, Easicode, STEVE and the Alphacode Tabulator have been produced at intervals at various DEUCE establishments, to all of which credit should be given for the introduction of new ideas and exploitation of older ones. A study of the way in which most of these schemes develop ideas in the others, and contribute new techniques in itself an interesting genealogical exercise. Extract: TIP TABULAR INTERPRETIVE PROGRAM (T.I.P.) This program devised at Bristol Aero Engines is the most remarkable example of simplicity in programming. The authors, faced with the problem of arousing computer consciousness in a number of engineers solved it in a novel way. DEUCE was presented to the potential users as something with which they were already quite familiar, namely, a desk machine operator with a fast machine and a huge sheet of paper filled into rows and columns. Instructions to DEUCE are then in a form exactly analogous to those to the human operator. We may, for instance, be accustomed to telling an operator to write successive integers in the first column, to square these numbers and write them in the second column and then subtract the numbers in the first column from those in the second. The Tabular Interpretive Program will accept these instructions in an even more concise form than they have been written above. Each instruction to T.I.P. comprises four numbers a, b, c and r. The number r tells the machine which of 31 operations it has to obey, and the numbers a, b and c detail the columns concerned. Thus incomplete program for performing the above operation. The first instruction has r = 4 which means 'read data into the column designated by c, in this case column 1'. The second instruction has r = 0 which is interpreted as 'multiply the numbers in column a (1 in this case) by the corresponding numbers in column b (also 1 in this case) and write the results in column c (i.e. column 2)'. The reader will deduce from the next instruction that r = 3 means 'subtract the numbers in column b from the corresponding numbers in column a and write the results in column c'. Similarly r = 5 is interpreted as 'print out results from column a'. The user has therefore at his disposal a 'human operator equipped with a sheet of paper ruled into 128 columns and 30 rows and with the facility of rubbing out any columns no longer required and writing new results there'. In the above program, for instance, the last two instructions could have been: a b c r2123 2005 causes each of the numbers in column 1 to be subtracted from PI'. The codeword a b c rNa b c 13 causes 'constant number a to be transferred to row b of column c' and a b c r a b Nc 13 causes 'the number in row b column a to be transferred to the cth constant store'. Two concepts which are strange to the newcomer to programming are those of counting the number of times round a loop and of modifying instructions. In these respects, simplicity is the essence of the T.I.P. facilities. Any value a, bore which has an asterisk against it is automatically increased by one every time it is obeyed. The instruction The order code is particularly well fitted to scientific and engineering calculations. Apart from the elementary arithmetic operations and 'read' and 'print', values of r have been allocated to logarithms and exponentials, trigonometrical and inverse trigonometrical functions, square roots, modulus, and sum series. The instruction r = 15 causes column specified in c to be filled with the progressive sum of the elements in column a. Other values of r are used for shifting a column of figures up or down by one row, this facility being of use in finite difference operations. Functions for interpolation among the elements of a column and for three-dimensional linear interpolation in data read to the machine by a special 'read' codeword are also available. Of course, in practice, it is frequently required to operate on a column not with the contents of another column but with a constant. For this purpose, 128 constant stores are available, 32 with built-in constants (which may, however, be over-written) and the remainder at the programmer's discretion. These stores are referred to as NO, Nl, etc., to N\21. Thus, the codeword: a b c r1 JV4 2 3 causes the constant JV4 which happens to be IT 'to be subtracted from each of the numbers in column 1 and the corresponding results written in column 2'. Similarly a b cAT4 1 '2 a b c r0 b c 16 or 0 Nb c 16 is interpreted as 'jump to instruction number c until the instructions from c onwards have been obeyed b (or Nb) times'. In these instructions b and c do not refer to column numbers. If b is specified the assumption is that the programmer knows precisely how many times he wishes to go round the loop (which probably contains some asterisked instructions); if Nb is used, this constant store may well contain a number (the number of times round the loop) which has itself been computed by the program. For example, the three instructions Instruction number a b c rc 1* 100 100 2-c + 1 0 99 c 16 c + 2 0 0 0 18 cause the first 100 columns to be added together. The instruction r = 18 which may only follow an r = 16 instruction causes all the asterisked instructions in the loop ending with previous instruction to be reset to Their initial value. Up to second order loops are permissible. There is one further instruction of this type for use with iterative loops: a b c r a Nb c 17 means 'jump to instruction c repeatedly until such time as the corresponding numbers in columns a and a + 1 are all within the percentage tolerance specified in the bth constant store, and then proceed normally'. A persistent annoyance to a programmer is to be held up by a failure instruction on a machine due to trying to divide by zero, or trying to find the square root of a negative number or other error due to the data or the logic. Such errors can be time-consuming on the machine, particularly when, as is the case with T.I.P. the programmer does not do his own testing. (Indeed the illusion of the machine as a piece of paper is maintained to the point that the programmer need never have heard of a punched card—the usual DEUCE input/output medium.) The present version of T.I.P. deals with such contingencies in two ways. If a tolerance only just fails—for instance, trying to find the inverse cosine of a number greater than unity but less than 1 + 2~25, T.I.P. gives the programmer the benefit of the doubt, and assumes that the data was badly rounded off. If, on the other hand, the data is quite impossible, the computer does not stop and involve a post-mortem and reference back to the programmer: it merely inserts a 'dash' in the corresponding result row. Any further operation on a 'dash' results in a 'dash,' and so the final results may well be quite satisfactory apart from one dash in a particular row which the programmer may easily be able to explain. The 'dash' technique has also been exploited to deal with discriminations. The operation r = 25 discriminates on the column of numbers specified by b, and where they are positive the corresponding numbers of column a are transferred to column c, and a 'dash' is placed in column c -f 1; otherwise a 'dash' is placed in column c and the number is transferred to column c + 1. Columns c and c + 1 can then be operated upon independently by the remaining program. Similarly columns can be merged by an r = 26 instruction. To avoid any scaling worries, and yet to preserve the speed of fixed-point working, all operations are carried out in block-floating arithmetic, that is, all the numbers in one column are stored to the same exponent, with the largest element filling the word and the exponent stored separately. Constants are stored as floating numbers. The number of elements in a column (usually a constant throughout a program, but varying if finite difference operations or interpolation in 'graphical' data are performed) is also stored with each column, as is the sum of all elements of the column. The latter is tested every time reference is made to the column. Up In 511 instructions can be stored in the machine at one time; in general (his is more than enough, but more can be read in, if necessary, to overwrite part of the program. Generous facilities for program testing are available, and the setting of specified keys on the machine will cause the machine to stop on a particular instruction, change it for another, or punch some (usually only the first) elements from all or specified columns. The latter limited punch out facility is of great use for the more involved errors. The organization of the program is fairly obvious. The 128 'columns' are in fact 128 tracks on the drum, the 30 'rows' are 30 of the 32 minor cycles per track, the remaining two being used for the number of rows, . the number of binary places and the sum check. The high speed store is used to contain as many of the subroutines for the various r functions as can be accommodated, and enough instructions to bring down any other selected r function. It also holds, of course, the columns currently being operated upon, and the routines for disentangling the a, b, c, r instructions and fetching and storing solumns. The scheme thus gives (in common with Alphacode) the opportunity of using DEUCE without having heard about punched cards, delay lines, minor cycles or magnetic drums, and one can begin to program for T.I.P. on the day one first hears of it. in Goodman, Richard (ed) "Annual Review in Automatic Programming "(1) 1960 Pergamon Press, Oxford view details Most of the papers are written in a style and manner which seem to have become universally accepted for papers on computer programming, at least in the English-speaking world and probably in others. This style insists on a liberal dosage of impressively detailed flow charts which, considering the well-known and understandable reluctance of programmers to read their own programs much less those of others, one suspects most readers hastily skip over, willingly granting their authenticity. The flow charts are invariably accompanied by long lists of special instructions described in the private patois of the author, who seems blissfully unaware or unconcerned that his specially constructed vocabulary of acronyms may present;. rough going to the reader from the inlying provinces. Finally, the style demands long and wearisome descriptions of basic concepts (e.g., subroutine; symbolic instruction, etc.) long since familiar to the average reader, some indication of difficulties as yet to be surmounted (e.g., automatic storage allocation; easier debugging; et al). Nevertheless, the volume does give some idea of the status of automatic programming systems in Great Britain in early 1959. It also contains a concise description of the 709 SHARE operating system, and another brief account of FLOW-MATIC and MATH-MATIC. There are two interesting appendices worthy of mention. Appendix One consists of reprints of two papers by the late A. M. Turing, "On Computable Numbers with an Application to the Entscheidungsproblem", in which the "Turing machine" was conceived, and a brief corrective note on the same subject. Appendix Two contains the "Preliminary Report of ~ ACM-GAMM Committee on an International Algebraic Language", since published elsewhere. The reviewer cannot suppress the question of whether this sort of material (Appendices excepted), so soon obsolescent or obsolete and so difficult to present adequately in short papers, deserves the effort and expense required to reproduce it between the bound hard covers of a handsome book. in ACM Computing Reviews 2(03) May-June 1961 view details DEUCE The English Electric DEUCE grew out of an active collaboration between English Electric and NPL. The DEUCE was based closely on the Pilot ACE (Haley i956). The initial software effort for the DEUCE lay in converting the existing Pilot ACE programs developed by NPL. Most of this work was done during 1955 in a combined effort between the users of the first three DEUCES, which were installed at English Electric, NPL, and the Royal Aircraft Establishment. This conversion work was in fact coordinated by NPL; it seems that in the mid-1950s English Electric did not see the provision of programming systems as part of their brief, although they did organize the DEUCE Users Group and a library service. Several active programming groups were associated with DEUCE installations, and by 1958 three important interpretive schemes for the DEUCE had emerged: GIP, TIP, and Alphacode (Robinson 1959). These three schemes had complementary domains of application: GIP was, of course, the famous matrix interpretive scheme from NPL, TIP was used for calculations on vectors, and Alphacode was used for scalars. The GIP matrix scheme was easily the most important programming system for DEUCE. Apart from its remarkably high speed, CIP was noted for its reliability. By means of check sums and other devices, complete confidence could be had in the results in spite of the inherent unreliability of the DEUCE (which had no parity checking, for example). The TIP (Tabular Interpretive Program) scheme was in effect a variant of GIP restricted to vector operations. The system was designed by the DEUCE group at Bristol Aero Engines to simplify programming for engineers and was widely used. TIP was a rather elegant system and required no formal understanding of linear algebra. It was intended to be accessible to anyone who was familiar with a "desk machine ... and a sheet of paper ruled into rows and columns" (Robinson 1959). TIP is an interesting relic of the transition from machine language to true programming languages. The third interpretive scheme, Alphacode, was specified by S. J. M. Denison of English Electric as an automatic coding system for naive users and for one-time jobs; Alphacode was directly inspired by the Manchester Mark I Autocode (Denison 1959). The interpreter produced programs that were typically about five times slower than conventionally coded programs, actually a considerable achievement considering the high speed of DEUCE when optimally coded. In November 1957 a project for an Alphacode translator (as opposed to the existing interpreter) was begun (Duncan and Huxtable 1961). The aim was the exceedingly ambitious one of producing translated programs as good as hand-coded ones. The translator was developed by F. G. Duncan, working at first with E. N. Hawkins and later with D. R. Huxtable. The system came into use toward the end of 1959. It was one of the most impressive programming achievements of its day, both in terms of sheer size (22,000 instructions) and in the difficulty of producing code for a machine with a multilevel store. The translator in practice produced code that was about two-thirds as good as handwritten code, a truly remarkable achievement given the complexity and subtlety of programming for the DEUCE. Several other programming schemes were produced for the DEUCE by other installations in the late 1950s. These included STAC, STEVE, GEORGE, SODA, and EASICODE (Robinson 1959). AH these systems were made available through the Users Group, but they do not appear to have been used as widely as the schemes already described. The development of software for DEUCE can be summarized as follows. The existence of a large amount of high-quality software from NPL led English Electric into believing that it was unnecessary to develop further programming systems. English Electric did see the need to coordinate and distribute programs through the Users Group and to organize programming courses. English Electric's failure to make a timely provision of an autorpatic programming system for DEUCE led to a number of ad hoc developments at various DEUCE installations during the period 1957-1959, which was a wasteful duplication of effort. In underwriting the Alphacode translator, however, English Electric demonstrated that it had at last come to recognize its duty to provide programming systems for the DEUCE. In January 1960 English Electric transferred its programming staff to the Data Processing and Control Systems Division at Kidsgrove, where an automatic programming section was established under the management of F. G. Duncan (1979). At this point, machines such as the KDF 9, for which excellent software was produced, were on the horizon. Extract: Conclusions Conclusions When we compare the development of programming at the three centers -- Cambridge, Manchester, and Teddington -- there are several factors to consider. First, we must consider the quality of the programming system; this is a subjective issue that ranges from the purely aesthetic to the severely practical -- for example, from the elegance of an implementation at one extreme to the speed of a matrix inversion at the other. We must also consider the failures of the three centers, especially the failure to devise a programming system that exploited the full potential of the hardware. Finally, we must consider the influence of the programming systems on other groups; this is less subjective -- it was described in the previous two sections and is summarized in Figure 2. Few could argue that Cambridge devised the best of the early programming systems. The work done by Wilkes and Wheeler stood out as a model of programming excellence. Cambridge made several outstanding contributions to early programming: the use of closed subroutines and parameters, the systematic organization of a subroutine library, interpretive routines, and the development of debugging routines. Perhaps the finest innovation was the use of a symbolic notation for programming, as opposed to the use of octal or some variant. It is difficult for us today to appreciate the originality of this concept. If Cambridge can be said to have had a failure, it was the failure to develop programming languages and autocodes during the middle and late 1950s, as reflected in the second edition of Wilkes, Wheeler, and Gill (1957), of which Hamming said in a review, It is perhaps inevitable that the second edition, though thoroughly revised, does not represent an equally great step forward, but it is actually disappointing to find that they are no longer at the forefront of theoretical coding. (Hamming 1958)] By neglecting research into programming languages, Cambridge forfeited its preeminence in the programming field. In the early 1950s, however, Cambridge was by far the most important influence on programming in Britain. This came about partly through the excellence of the programming system and partly through the efforts that Cambridge made to promote its ideas. Two machines (I`EO and TREAC) based their programming system directly on EDSAC, and five machines (Nicholas, the Elliott 401 and 402, MOSAIC, and Pegasus) were strongly influenced by it. It is also probably true that no programming group was entirely uninfluenced by the Cambridge work. Overseas, the influence of the EDSAC programming system was just as great, largely through the classic programming textbook by Wilkes, Wheeler, and Gill (1951) (see Campbell-Kelly 1980a). At Manchester the programming system devised by Turing for the Mark I makes a disappointing contrast with the elegance of the Cambridge work. From the point of view of notation, it is difficult to find a single redeeming feature. Probably the only feature of real merit was the concept of dividing a program into physical and logical pages. Echoes of this idea can be discerned in today's segmented computers. In its way, Turing's programming system did have considerable influence, for all efforts to replace it with something more suitable were curiously unsuccessful. Thus programmers for both Mark Is and all seven Mark Iota's had to struggle with Turing's clumsy teleprinter notation throughout the life of these machines. Here is perhaps one of the most valuable lessons of this study: poor design decisions taken early on are almost impossible to correct later. Thus even when people with a Cambridge background arrived at Manchester, they were unable to make a really fresh start. By producing two successive input routines that were not much better than Turing's, they managed to combine the worst of both worlds: an unsatisfactory programming system that was not even a stable one. The one real high spot of the Manchester programming activity was Brooker's Mark I Autocode. Brooker's achievement was the most important programming event of the mid-1950s in Britain. If Brooker had not devised his autocode at that time, programming in Britain might have developed very differently. The autocodes for DEUCE and Pegasus were directly inspired by Brooker's and had considerable notational similarities with it. Beyond the time scale of this paper, Brooker's Mark I Autocode and his later Mercury Autocode (1958) were a dominant influence on British programming until well into the 1960s, when languages such as ALGOL 60 and FORTRAN came onto the scene in Britain. Of the three programming systems devised at Cambridge, Manchester, and Teddington, it is probably the latter that inspires the least passion. Ii the punching of programs in pure binary was an efficient method, it was also a singularly uninspiring one. Curiously, aficionados of the Pilot ACE and the DEUCE had great enthusiasm for programming these machines, which really had more to do with the joys of optimum coding and exploiting the eccentric architecture than with any merits of the programming system. In many ways the crudity of the programming system for the Pilot ACE was understandable: the speed of events, the lack of a backing store, and so on. But perpetuating it on the DEUCE was a minor tragedy; by replicating the programming system on the 32 commercially manufactured DEUCES, literally hundreds of rank-and-file programmers were imbued in this poor style of programming. MOSAIC (Section 3.4) shows that it was entirely possible to devise a satisfactory programming system for machines of the ACE pattern; it is most unfortunate that this work was not well enough known to influence events. NPL did, however, have one notable programming-success: the GIP matrix scheme devised by Woodger and Munday. This scheme became the sole reason for the existence of many DEUCES. The reliability of the mathematical programs produced by NPL, their comprehensiveness, and their speed have become almost legendary. A history of numerical methods in Britain would no doubt reveal the true role of NPL in establishing the methods of linear algebra as an analytical tool for the engineer. In an interview, P. M. Woodward, one of the principals of the TREAC programming activity, recalled, "Our impression was that Cambridge mattered in software whereas Manchester mattered in hardware" (Woodward and Jenkins 1977). He might well have added that NPL mattered in numerical methods. Because this paper has been primarily concerned with the development of programming during the period 1945-1955, Cambridge has received pride of place as the leading innovator. Had the paper been concerned principally with hardware or numerical methods, however, the ranking of the three centers would have been different. But considered purely as innovators of programming, there can be no question that Cambridge stood well above the rest. Abstract: By 1950 there were three influential centers of programming in Britain where working computers had been constructed: Cambridge University (the EDSAC), Manchester University (the Mark I), and the National Physical Laboratory (the Pilot ACE). At each of these centers a distinctive style of programming evolved, largely independently of the others. This paper describes how the three schools of programming influenced programming for the other stored-program computers constructed in Britain up to the year 1955. These machines included several prototype and research computers, as well as five commercially manufactured machines. The paper concludes with a comparative assessment of the three schools of programming. in Annals of the History of Computing 4(2) April 1982 IEEE view details Resources
|