RegneCentralen ALGOL(ID:3488/reg005)Danish family of Algol 60 languagesRegnecentralen was a semi-academic computer designer and manufacturer established by the Academy of Technical Sciences (Akademiet for de Tekniske Videnskaber, ATV) in 1955. They developed a series of ALGOL 60 dialects named for the computers for which they were made. The second of these, Peter Naur's GIER ALGOL, was a design masterpiece and hugely influential in the history of programming These details come from "RC4000, RC8000 and RC9000" below Algol 1 was DASK ALGOL Algol 2 - Algol 4 was GIER ALGOL Algol 5 was for the RC4000 Algol 6 (appr. 1970) featured "field" data types and records, and the "long" data type (48-bit integer). Algol 7 extended the language syntax with while-statements and repeat-until loops, and introduced "context records", a mechanism to support multi-user applications. Algol 8 introduced support for real coroutine programming with the "activity" concept, including separate stacks and a very nice integration with the Algol I/O system. The current version is called "Algol 8 Version 2". Many other features and enhancements than those mentioned above has been added over the years, and the run-time library has grown similarly. Structures: Related languages
References: Introduction The European ALGOL situation is a complicated one, made up of the influence of workers from about 25 institutions in ten countries. A brief account of the work done in all these places must emphasize certain developments and forget about others. However, even if in this way I run the risk of doing injustice to valuable work in one place or another, either from ignorance or because of the point of view I am going to present, I feel justified in attempting the present task because, after all, there is a great underlying unity of purpose in the European ALGOL effort. Differences of opinion that do exist concern the proper means of achieving the goal. About the goal itself there is hardly any disagreement. The plan of presentation of this paper is as follows: First I will try to describe the work done in certain distinguishable groups of centers, covering both historical development and the current emphasis within each group. When I have in this way covered the complete field I will try to single out the underlying basic points of view. This will bring me to my conclusion. Extract: The Central European Group (ALCOR) The Central European Group (ALCOR) The first group, both historically and with respect to the number of centers involved, is the central European one. It evolved around Zurich and Mainz and contains the roots of both ALGOL itself and of the whole idea of an algorithmic language. In fact, Rutishauser of Zurich is sometimes referred to as the "grandfather of automatic programming" in view of his work in this direction from about 1950.(1) Again the suggestion of a cross-Atlantic agreement in this field goes back to this group; of the four European members of the first ALGOL committee three came from these two places, namely Rutishauser from Ziirich and Bauer and Samelson from Mainz. The fourth was Bottenbruch from Darmstadt, a center which also belongs to the central European group.(2) It is, I think, worthwhile to notice that these four all come from universities, carrying a heavy load of mathematics and numerical analysis with them. Also, the machines they had available at the time when this work was begun were of moderate size and were used primarily for teaching and research. The meaning of the central European group cannot be understood, however, unless one knows about the idea of a more local, and far more rigid, collaboration which developed simultaneously with the international ALGOL deliberations. This idea again originated at Mainz and goes as follows: When implementing the common language on different computers it is likely that the details of the language will not be treated exactly alike. Consequently a direct interchange of programs will not work. In order to achieve strict interchangeability it is necessary to make the different translators essentially alike, not only with respect to the language they process, but also in their inner workings. This, then, is the idea of the so-called ALCOR group of computation centers: Within the group all members will make use of the translation algorithm developed in Mainz by Bauer and Samelson(3) and will have the actual matrix, which controls this algorithm, supplied from Mainz. In this manner complete uniformity of translation is assured. The ALCOR group (ALCOR stands for Algol COnverteR) at present consists of the following centers: Institut fur Angewandte Mathematik, Universitat, Mainz. Institut fur Praktische Mathematik, Technische Hochschule, Darmstadt. Zentral-Laboratorium, Siemens und Halske, Miinchen. Institut fur Angewandte Mathematik, Universitat, Bonn. Rechenzentrum der Technischen Hochschule, Munchen. Institut fur Angewandte Mathematik, Eidgenossische Technische Hochschule, Zurich. Institut fur Schwachstromtechnik, Technische Hochschule, Wien. Telefunken GmbH., Backnang. Standard Elektrik Lorenz, Stuttgart-Zuffenhausen. Zuse KG, Bad Hersfeld. Regnecentralen, Copenhagen. Oak Ridge National Laboratory, Oak Ridge, Tennessee. The ALCOR group was established in 1959 with the expectation that working translators would be completed within each of the centers within a year or so. However, the development of ALGOL 60(6) naturally caused some delay in the whole plan and the status of these centers with respect to the completion of actual compilers is as follows: Since the end of 1958 a translator for a subset of ALGOL for the machine Zuse 222 has been running. The subset excludes procedures. May 1961 saw the completion of an extended system conforming to the so-called ALCOR specifications. These include most of ALGOL 60, with the exception of recursive procedures and own. 2. SIEMENS UND HALSKE, MUNCHEN In November 1960 a translator for a modest subset of ALGOL for the Siemens 2002 was completed. The subset excludes procedures, arrays of variable size, and - own. A translator for the complete ALCOR specificstions for the same machine is well underway. 3. TECHNISCHE HOCHSCHULE, MUNCHEN Since February 1959 a translator for a small subset of ALGOL for the machine PERM has been working. By September 1961 a translator for a larger subset was completed. Work on an even more complete system is in progress. 4. ZURICH Since July I959 a translator for a small subset of ALGOL for the machine ERMETH has been working. During 1960 this was extended to include the complete ALCOR subset. This has, as far as I know, been working for some time. The work in Copenhagen will be described below. Oak Ridge falls outside the scope of this report. As to the remaining centers, I regret to say that up-to-date information has not been received. This successful construction of translators within the ALCOR group has nowhere been considered an experiment or a goal in itself, but has been the basic means of communication with the computer. Thus, at Mainz the teaching of the use of the machine and of numerical analysis has for several years been based exclusively on ALGOL. Some 400 people have taken courses in ALGOL, and all programs for the machine are written in this language. In fact, only one or two people in the center know the machine code of the Zuse 222. In this context it is also of interest that the Mainz translator for the Zuse 222 already has been distributed to eight other installations having this machine and that several dozen more are expecting it. As the last field of activity of the central European group, the Taschenbuch project should be mentioned. This project aims at the publication of a four-volume handbook of tested algorithms from the field of numerical analysis. The algorithms will all be written in ALGOL. In compiling this collection the collaboration of outstanding workers from all over the world has been sought. Also, it is intended that only algorithms which have been run successfully on at least two different computers will be included. Thus, hopefully, the handbook will present an all-round collection of algorithms which are sound both theoretically and practically. Summarizing, we see that the central European group has been busy in a wide field of practical implementation and use of ALGOL. Their translators have initially been modest; often quite heavy restrictions of the language have been tolerated. In this way the actual work of constructing the translators has often been kept very low. In fact, figures - the range from 1/6 to $5 man years are quoted. These figures of course do not include the general advisors of the group. At the same time the group has been very active in the international work of defining the language and is unquestionably responsible for a great deal of the basic characteristics of the language. Extract: The Dutch Group The Dutch Group All other European groups came into the ALGOL work when this was already well underway in central Europe. These groups include a Dutch group, a Scandinavian group, and some additional computation centers which do not fit into any grouping. Speaking of a Dutch group is perhaps going too far, since this would consist only of two centers: the Mathematical Center in Amsterdam and the Dr. Neher Laboratory of the Dutch Postal and Telecommunications Services in Leidschendam. These two centers, although keeping in close touch with each other's developments, do not have any formal arrangements between them. Even so, there is so great a similarity in the background and approach of these centers that it is convenient to pool them. Whereas the central European group is dominated by people with a background of applied mathematics and numerical analysis, the Dutch group takes its direction from people who have a taste for logic and formal analysis and who, in addition, have been directly concerned with the design of machines. Thus, van der Poel, at Leidschendam, is known for his analysis of the simplest possible computers and his contributions to the design of the ZEBRA, while the Mathematical Center at Amsterdam has been the cradle of both the ARMAC and the Electrologica X1. These characteristics have left their mark both on ALGOL 60 and on the translators designed by the groups. Both groups were very active in the work leading up to ALGOL 60, and van Wijngaarden from Amsterdam sat on the final thirteen-man committee.(6) In this work the Dutch would urge generality and removal of arbitrary restrictions in the language and also contributed several original features to the language. In the field of translator construction the group at Amsterdam has been singularly successful. In keeping with their general outlook they have strived to produce a translator which conforms absolutely with the official definition of the ALGOL 60 language, with as few restrictions as possible. This has brought about a new attitude towards the problem of processor constructions, the primary focus being the running program, while the actual translation process, the transformation of the input string, becomes a relatively minor matter.(4, 5) A system designed along these lines was constructed at Amsterdam by two people, Dijkstra and Zonneveld, and was first tested in June 1960, i.e., about 5 months after the language was defined. This is not only the first successful system that will handle the full generality of ALGOL 60 procedures, but probably remains the only one which will handle recursive procedures. The only restriction is dynamic arrays. At the Dr. Neher Laboratory an ALGOL system which goes to even higher degrees of conforming to the letter of the ALGOL 60-report is underway. To most people who have worked with implementation of ALGOL 60, the idea of a system that handles the complete language is likely to provoke the question: What about the efficiency? This is indeed one of the most freq;ent objections raised against ALGOL 60. It is too general, and therefore-inefficient and difficult to implement, people will say. In view of this reaction it is interesting to analyze the attitude and practical experience of the Amsterdam group. First, on the question of attitude, I want to cite some reflections made by Dijkstra(4). Dijkstra says that the approach taken will clearly cause some slowing down of the running program. However, as a scientific institution, the Mathematical center in Amsterdam prefers to work on a programming technique that will gain in importance in the immediate future. And there are good reasons to believe that the trend of machine design will be such that a proportionally greater number of machines will come into existence for which the cost of a very general approach is not excessive. As a second point Dijkstra considers that the increased flexibility and the removal of arbitrary restraints to be taken into account in will be found more and more valuable as time goes on. Indeed, for the utilization of the rapidly increasing machine capacity these points may well turn out to be decisive. On the practical side it should perhaps be said first of all that the computer at the Mathematical Center in Amsterdam, the Electrologica X1, happens to be rather well suited to the general approach taken. Even so, it is significant that the ALGOL system does not remain an academic exercise. On the contrary, the use of it has been taught to a great number of people and it is being used extensively for solving a wide class of practical problems. Extract: The Scandinavian Group The Scandinavian Group The Scandinavian group consists of three centers in Sweden, namely Matematikmaskinnamnden in Stockholm, FACIT in Gothenburg, and Svenska Aeroplan Aktiebolaget in Linkoping, and one center in Denmark, namely Regnecentralen in Copenhagen. The latter is my own institution. Of these institutions one is a machine manufacturer and service center, one is a user, while the remaining two are combined service centers, machine designing and building groups, and research institutions. It may seem strange that these rather different backgrounds should provide an incentive to close collaboration, and probably the close contact between our groups is due mainly to a general positive wish to share our results with any interested group. In 1954 the group at Stockholm very generously put the design of their machine BESK at the disposal of all of us. Like the Dutch group we came into the work only after the preliminary specifications of ALGOL had been settled, but early enough to become quite heavily engaged in the work of establishing ALGOL 60. Throughout this work our attitude has been that we would be willing to accept any feature in the language on which it would be possible to agree. In fact, we have felt that the language would in any case, be so much better than anything we could think up ourselves, and since we needed a language at the time ALGOL was first announced, we accepted it with glee. The fact that in addition ALGOL seemed to have a fair chance of becoming a widely adopted standard, of course, added to our satisfaction. With this attitude I believe it is only natural that we should want to help further the development in whatever area we could be most useful, and our work in trying to organize and facilitate free discussion and the interchange of ideas is one fairly obvious response to this desire. This work was started in March 1959 and was initially planned to cover only the communication within the ALCOR group. However, by June of the same year we had already agreed to act as a general clearing house for communications from any interested person in Europe. As some of you know, we have carried out this assignment through the medium of a series of duplicated notes known as the ALGOL Bulletin. These are circulated free of charge to anyone who is interested in getting them. The material in the Bulletin consists entirely of the contributions from the readers. So far, in fact, no material submitted has been rejected. At the same time a considerable amount of work on implementation has been going on, although so far with rather less success than in central Europe or in Holland. In fact, the first Scandinavian translators in Copenhagen and Gothenburg have only just started running. The one in Copenhagen has a power about equal to that of the ALCOR specifications, i.e., it will not handle recursive procedures or own arrays(7). The one in Gothenburg is somewhat more limited. Extract: Independent Centers Independent Centers Of projects falling outside the three groups I have described so far, I would like to mention developments at Zeiss, Jena, Germany, where they are working on a translator for the Zeiss-Rechen-Automat ZRA 1, and at English Electric, Kidsgrove, England, where a translator for a new machine, the KDF 9, is being developed. At this point I would also like to make clear that since this report covers only Europe the developments in Japan and Asian Soviet Union are not touched. Extract: General Trends General Trends Having now described the European ALGOL situation in a geographical and historical manner I would finally like to try to find some of the underlying trends and points of view. I think it is right to say, first of all, that throughout Europe ALGOL has been taken very seriously. People have put a great deal of thought, work, and faith into the whole idea. However, underneath this common general faith we can discern at least two different attitudes. These can perhaps be expressed as follows: in the one view ALGOL is seen as the final expression of existing trends and needs within the fields of numerical analysis and machine design. Where ALGOL goes further than these it is regarded with suspicion or rejected altogether. In particular, where ALGOL forces an inconvenient use of present machines it is restricted in the actual implementations. In the other view ALGOL is seen as an independent construction in its own right. Where it provides new mechanisnk for which no immediate use is apparent, this is understood as just a temporary state of affairs, the conviction being that these features will turn out to be important as soon as the learn to use them. Again, if ALGOL forces an inconvenient use of present machines, this is thought to be the fault of the machines. Indeed, it is felt that future machine design will be heavily influenced by ALGOL. If you ask for my opinion on these two attitudes, I must say that I think that they are both right and useful. Unquestionably there is a big job to be done in working out the best algorithms of numerical analysis in a common language like ALGOL, and for this purpose people probably can do without the most general features of the language, at least to start with. At the same time I feel that there is an urgent need for radically new impulses in the logical design of machines. Throughout the past decade machines have been designed according to roughly the same principles--or lack of principles. The machine coding aspect has always been in the foreground in spite of the fact that the inconveniences of present-day machine code are obvious and widely recognized. Previous languages, such as FORTRAN, have not brought about a change here, basically because these languages have themselves been designed with the computer in mind. Thus, they have only tended to perpetuate the present machine designs. This is of course fine, if one feels that it is right to spend, first, a great effort in building a machine which in a sense no one will use, and then another great effort in writing a system of 20,000 or 50,000 instructions necessaryto work with the thing. I feel that there is something utterly wrong here, and I am gratified to see that apparently the only incentive to changing this situation is corning from the ALGOL effort. Extract: Conclusion Conclusion In conclusion I think it is fair to say that within a significant part of the European computation centers the ALGOL effort has had a profound effect. It has inspired wide interest in the fields of computer languages and translation. It has furthered the contact between these computation centers. It has convinced a great many people that a common language is both desirable and possible. It has inspired new thought on the machine design problem. Finally, it has led to the completion, in many places, of processors of a high degree of generality and usefulness. in Proceedings of the 1961 Computer Applications Symposium, Armour Research Foundation, Illinois Institute of Technology, Chicago, Illinois view details in Proceedings of the 1961 Computer Applications Symposium, Armour Research Foundation, Illinois Institute of Technology, Chicago, Illinois view details |