2007: Cecilia Berdichevsky: The Beginning of Computer Science in Argentina.

The Beginning of Computer Science in Argentina-Clementina-(1961-1966)

Cecilia Berdichevsky

I have the honor of dedicating this work to the memory of Dr. Manuel Sadosky, my mentor and friend.

Abstract

1957 marked the beginning of the era of modern computing in Education in Argentina. I was lucky enough to be a living part of this history. That year, after an international bid prepared by a Commission of the University whose members unanimously voted for the Mercury of Ferranti computer the machine was purchased by the University of Buenos Aires (UBA). It was installed in 1961, and an Institute of Calculus was created with the aim of improving its professional and technical applications. Almost at the same time, a new university career was organized: Computer Science, and those three events, promoted by our teacher and mentor Manuel Sadosky, marked the beginning of the era of computers in education in our country. Work at the Institute covered three fields: problem solving, research and teaching. Several Working Groups were organized with the aim of solving “real problems" in different disciplines: Mathematical Economics, Operations Research, Statistics, Linguistics, Applied Mechanics, Numerical Analysis, Electronic Engineering and Programming Systems. The architecture, structure, operation, languages and other characteristics of the machine, quite advanced for the time, determined the chosen areas. After the military coup of 1966 that took over several institutions, part of the University and several of its work teams, especially our Institute of Calculus, the era of Mercury in our country came to an abrupt end, ceasing all operations shortly after. 90% of the members of the Institute, scientists, professors and highly trained professionals, resigned and many of them left the country taking their knowledge and expertise abroad, and all of us, those who left and those who stayed, had sound successes in everything we undertook afterwards.

1  Introduction

1956, when I started my studies for a degree in Maths, some leading personalities from several universities in Argentina were thinking about the idea of becoming a part of the modern computing world. One of the actions was the creation of a Calculus Institute in the UBA with the aim of promoting the development of Applied Mathematics using resources of Electronic Automatic Computing. In 1957, a public international bid was prepared by a Commission of the UBA created with the express purpose of providing the School of Natural and Exact Sciences with a modern computer. With that purpose a special subsidy was provided by the CONICET (National Council of Scientific and Technological Research). As for the biding, four companies presented their proposals: IBM, Remington and Phil co from the USA and Ferranti from England. After a careful study, and taking into account the technical characteristics and prices, the members of the Committee unanimously recommended the purchase of the Mercury offered by the English company Ferranti from Manchester. It was in those days a state-of-the-art machine with excellent technical features: speed, types of memory and mainly because Autocode, the language developed for Mercury by the University of Manchester researchers was easy to learn, user friendly for the time, and fit for dealing with scientific problems. In 1960, when the computer arrived in Buenos Aires, and was installed a year later in the environment were I was working, I was immediately attracted by the computing work. From the beginning we named our machine “Clementina", because it was programmed to play the popular American melody “oh my darling Clementine", that was easily recognized even if the machine sounded in a peculiar way. Some time later we programmed it to play opera songs and a known tango named La Cumparsita. 1957 to 1961 were devoted to the training of the first group of analysts and programmers, and to informing possible future users of the news of the arrival of a modern computer at the University. That was one of the landmark experiences in the development of computing in Education in the country.

2  Installation of the computer in the University of Buenos Aires

The installation of Mercury was completed at the beginning of 1961. The reason for the delay was that the room being prepared on the second floor of the new building of the School of Exact and Natural Sciences was not ready and did not meet the strict Ferranti specifications yet. The staff members appointed to work with our Mercury where going to be trained in Manchester, in the Ferranti plant. This was one of the factors that tipped the balance of the choice of Mercury. Actually, training had begun in Buenos Aires by Cicely Popplewell, who had worked with Alan Turing. Ernesto Garcia Camarero, a Spanish mathematician engaged by the University with the purpose of helping in the training of the future users, also participated in the formation of the annalists and programmers of the Institute.

3  The Calculus Institute and its Working Teams

With the aim of doing research, working with the machine, learning, teaching and using it with different purposes, an Institute of Calculus was created by the University at the end of 1962, and a new career, called Scientific Computist, was organized, both promoted by our mentor, teacher and leader Manuel Sadosky, who has also been instrumental in the decisions and actions that led to the era of Scientific Computing in Argentina. One of the tasks of the Institute was tackling “real problems", and thus two lines of work where defined: on one hand the Institute received and attended problems to be processed and solved, from researchers and professors of the different UBA Schools or other national institutions that needed its services. On the other hand, different specialized Work Teams were created to work on different subjects M.:

 Mathematical Economics: That group was the largest one and was a multi-disciplinary team conformed by economists, sociologists, statisticians and professionals from other disciplines. It was directed by Oscar Varsavsky, and integrated by Arturo O’ Connell, who was and is an important economist, Jorge Sabato, later a Minister of Education, Victor Yohai, a statistician, and many others. This team produced two economical models Meic-0 and Meic-1, developing a new technique, which used the computer to process statistical data provided by Argentinean sources.

 Operation Research: Began working on a problem of great national relevance: the study of the use of rivers of the Andes ridge, using numerical models. Two national institutions proposed the project: the Federal Council of Investments (CFI) and the CEPAL. An automatic “rolling mill" for the company Siderca, and a Human Diet for CONADE (The National Council for Development) were also performed. That team was directed by O. Varsavsky and Julian Aráoz, with the advice of renowned engineers and university teachers, like Roque Carranza, who was a professor in the Department of Mathematics , and later a Minister and President of the CONADE. The work of this group became one of the first examples in the world of the application of numerical experimental methods in complex dynamical systems. This same group directed by Julian Aráoz and conformed between others, by Juan Carlos Fränkel, Marcelo Larramendy and Nestor Sameghini carried out the first computer works on Pert, Linear Programming and Critical Path Analysis.

 Statistics: One of the main objectives of the Institute was to promote the study and application of Statistics, as yet insufficiently developed in the country. The Statistics team directed by Sigfrido Mazza worked in two levels: on one hand, special studies were undertaken about the problems concerning the work of important national institutions, like the National Institute for Agricultural Technology (INTA), the National Petroleum Company (YPF), the National Telephone Company, and the National Health Institute, among others. The group also had the responsibility of designing the sample and evaluating the errors of the compiled material of the 1960 Population Census. On the other hand, part of this team centered its activities on the permanent collaboration with INTA and published a work about “The use of Mercury Ferranti computer in the analysis of experimental data", to inform users all over the country on how to normalize the compilation of statistical agricultural material.

 Applied Mechanics: This group was directed by Mario Gradowczyck with the collaboration of a team composed mainly by engineers like Jaime Schujman. That group had two lines of work: mechanics of solids and mechanics of fluids and structural calculus for local companies of engineering.

 Numerical Analysis: this team was formed under the direction of Pedro Zadunaisky, a well known scientist in Planetary Mechanics and the participation of Victor Pereyra, Enrique Ruspini and others. A team of Numerical Analysis was created. I had the privilege to be one of its members, working in problems of convergence in the numerical solution of differential equations regarding the calculus of planetary orbits especially that of the Halley Comet to which Zadunaisky devoted a great part of his life and had worked on at the Smithsonian Institute. As an acknowledgment for his work an asteroid is named after him.

 Computational Linguistics: was a section directed by engineer Eugenia Fisher integrated between others by Victoria Bajar who developed an important career in Mexican university institutions. The subject of that team was automatic translation especially from Russian to Spanish and vice versa, and the structure of Spanish in collaboration with the cathedra of Philology of the School of Philosophy and Literature of the UBA, and other institutions. They had a communications sector in which Juan Carlos Angió performed works for ENTEL (the National Telecommunication Company).

 Electronic Engineering: was a very particular team directed by Jonas Paiuk who was trained in Manchester and became the chief of Maintenance of Mercury and Oscar Matiussi followed in the direction of the team. Their main task was to ensure the proper functioning of the computer, something they accomplished with exceptional efficacy. They also undertook Research and Development tasks to improve the equipment, mainly its input and output, i.e. the construction of a converter from punched cards to punched ribbon, installation of new magnetic drums, a line printer and also an analogical to digital converter whose destination was the analysis of neurological data required by a research group at the Buenos Aires Children Hospital.

 Programming Systems: This team was established under the direction of Wilfred Duran. Its main achievement was the creation of a new language the COMIC (Compiler of the Calculus Institute). See the details of this project in 4.3.Durand.

4  Working in the Institute of Calculus

4.1  Fields of work:

Work in the Institute, covered three fields:

1.  Solving problems on demand by scientists and professors of the different UBA Schools, other Universities of the country and public and private institutions of all kinds.

2.  Application of the computer to research in the different fields the scientists of the Institute worked in.

3.  Teaching how to program and use the computer and the languages that it allowed, a task every one of us took part in. Some of us also lectured on different subjects in careers related to computing or to the Institute, especially those enrolled in the new career of Scientific Computist.

4.2  Work at the Institute:

At first, the team was quite small, only 7 or 8 people including our leader and mentor, Manuel Sadosky and Rebeca Guber a mathematician, and excellent organizer and public relations person, who was Chief of Services during the entire life of the institution, and managed the Institute’s every days operation, problems, services and financial concerns. By the time Mercury arrived and operations began, I had the luck of having had in hand a real problem of Physics that, Mercury solved as soon it was installed.. In one of her first mornings in Buenos Aires, Cicely Popplewell gave me a private lesson in which I got instructed of the programming needs of the problem to which I had devoted several months of Nestler Rule calculation with not reaching the solution. That afternoon, the problem got solved. Like my colleagues, I also helped in solving problems that were the everyday worries of different researchers of some universities of the country and other national institutions, like the Departments of Physics, Meteorology, Chemistry, Industries, and others from our UBA Schools, like the School of Engineering of the UBA, National Universities of La Plata, Cordoba, Rosario, San Juan., the Institute of Physics of Bariloche of the University of Cuyo, and even some Universities of abroad , like the University of Montevideo Uruguay, University of Chile and an Institute of Mathematics from Dublin, Ireland. National institutions like the National Commission of Atomic Energy (CNEA), the Meteorological National Service, the National Company of Communications and others, and some outstanding private companies like Shell, Ducilo, Mellor Goodwin and so on.

4.3  Working with Mercury

Work with Mercury was defined by its resources and its characteristics, structure and operational capabilities, as well as by the languages, routines, stored libraries and facilities that it offered

4.3.1  Characteristics of Mercury

Clementina was an improved version of the Mercury developed in 1955 at Manchester University. It was defined in the Manchester Manual as an Electronic Digital Computer of Stored Program, Large Size, High Speed and Scientific, which meant:

Digital computer:  Discrete representation, essentially counting arithmetic, direct descendant of the abacus and of counting with hand fingers. Besides counting it also had to perform functions that were not arithmetic such as: storing, retrieving data or instructions from memory, using intermediate results or organizational tasks. Those functions constituted the "red tape functions".

Stored program computer:  To be run, the program and data had to be completely stored in the machines memory and the instructions were executed one by one.

Large size computer:  The Mercury could operate at its normal speed only if it had instructions, operators and data stored beforehand in the memory. Therefore it needed a great space of memory. It also needed space to store the intermediate results, otherwise it lost speed.

High speed computer: Additions and subtractions, took 180 microseconds each and multiplication 300 microseconds. At that time, that was high speed indeed.

Scientific computer:   Mercury was considered a scientific computer unlike data processing or general-purpose machines, because it could perform large and complex numerical calculus of nuclear physics or aeronautic engineering, carry out a numerical integration of complicated functions of more than one variable, play chess, or prove theorems of symmetry logic. Nevertheless, it could also perform the daily routine of book keeping of a bank.

4.3.2  Some Features

Mercury was a first generation computer: because it worked on valves. To program in Autocode, and that was one of the main factors in favor of this language at the time, it was not necessary to know either the machine language or the structure or details of the "real machine", but the programming language and the running of the "ideal machine", that could understand the instructions.

4.3.3  Operation

The machines operation had concepts that were developed much further, 10 or 15 years, in other computers. For example, the pagination system was in Mercury by software and gave birth to the idea of pagination by hardware. Mercury was state-of-the-art for the time. Consider: the facility of pagination, selection of rounding or truncate to improve results, the use of sub indexes, cycles with negative steps and more, none of them used by other computers of the time. The machine had a physical structure fit for scientific tasks, so its capacity of calculus and processing speed was not too bad, but input and output operations were very slow.

4.3.4  Languages

By the time it was received, Mercury operated in three languages

1.  Absolute or machine language.

2.  PIG2, a symbolic language that was an Assembler.

3.  Autocode, a higher level language which was a Compiler, developed by A.Brooker from the University of Manchester and improved in 1957 for the Ferranti Mercury. It was the language used, until COMIC came along Broker R.A..

4.3.5  Basic Functions

The machine contained a set of basic functions and each step of the programming would be the use of one of them. Mercury could not perform more than one operation at a time, and they were the three basic arithmetical operations: addition, subtraction and multiplication. Division was not a basic function: the quotient of a pair of numbers consisted in executing a sequence of the 3 basic operations: a succession of subtractions, combined with the counting of the quantity of subtractions. In spite of the fact that multiplication was included in the basic operations, a repeated and counted addition could be enough, but being a scientific computer, Mercury had a set of electronic functions one of which was multiplication; in that way it took less than half the time than the iterated addition.

4.3.6  Input and Output

Input was done by a photo electrical reader that read the punched paper ribbon and the program ran instruction by instruction, only one at a time. The output after processing was produced by a paper ribbon perforator and teletype.

4.3.7  COMIC (Compiler of the Calculus Institute)

When the nature of the problems that were analyzed, studied and solved in the Institute became more complex, it was necessary to think of researching to create new languages to obtain the maximum potentiality from our equipment.

As it was said above, the team of Programming Systems under the direction of Wilfred Duran created the COMIC acronym for Compiler of the Calculus Institute. It was published in May 1966 and from the first moment it proved suitable for dealing with some of the Institute programming problems. The language was mainly created to satisfy the needs of the Mathematical Economics team directed by Oscar Varsavsky. The requirements of the Economical Models programs exceeded the Autocode capacity. With Comic, the variable identifiers got more length, and the language became more “user friendly". It also had additional operations to manage matrices and vectors. The use of Comic, had also the purpose it fulfilled perfectly, of facilitating the use of the machine to programmers from different institutions, without them having to be familiar either with the internal structure of the computer or with the machine language. Comic was constantly improved by adding new capabilities to it. Among those improvements were routines prepared by the students of the Career of Scientific Computist as part of their curriculum. For instance, Cristina Zoltan, one of the first Scientific Computists graduated, later an important authority and professor of the Simón Bolívar University in Venezuela, designed a routine for the use of the Graphic device that was purchased by the Institute at that time.

5  My Training abroad

In 1961, the International Computation Center with headquarters in Rome offered two fellowships to the UBA. Argentina was a member of the ICC and Dr.Sadosky the Argentine representative. I got one of them. The second was not used, because at the time nobody filled the required qualifications. The fellowship consisted of a six months stay at the University of London Computer Unit, the English equivalent of the Argentine Institute of Calculus and another six months at the Center of Nuclear Studies of Saclay, France which had an Arithmetic Electronic section. Both institutions had a Ferranti Mercury.

5.1  Stay at the London University Computer Unit

The issues I worked on in the Unit, whose one and only computer was a Mercury, were:

Programming Techniques: In the Unit, I got acquainted with a very powerful new method of calculating the eigenvalues of a general matrix, the Francis method, which I studied, programmed and wrote the specifications of, under the guidance of the Unit staff. The program that calculated the eigenvalues of a general matrix up to the order 15x15 was successfully applied. The second problem, a library routine in which I was working in was also finished and became the 3rd. library routine for the Mercury of the Unit Thanks to the fact that the London Unit operated as an "open store", i.e. that the persons that knew how to operate it were allowed to use the machine, I could personally carry out all the tests, something that gave me a good background to operate our Mercury back in Buenos Aires.

Lectures and seminars:  During my stay in London, a weekly meeting took place in order to analise and discuss the CHLF 3 compiler, which was a new Input Routine. That seminar was very instructive; it gave me a insight into the structure and characteristics of a compiler. The Routine was finished and put to use at the Unit in April 1962. With the purpose of making possible the use of CHLF in our Institute´s Mercury, a new version, "version B", of the CHLF routine was written and with some adjustments made by members of the staff of our Institute, was put to use in our Mercury. I attended lectures in Numerical Analysis and a course in Matrix Calculus, by Professor Wilkinson. The second course in Numerical Analysis was developed by Professor Crank and the subject was: Numerical Methods for the solution of Partial Differential Equations allowing me some insights into new unpublished works.

Other English computer centers, such us the computing laboratory of Manchester University and the University of Cambridge were visited with the same purpose of analising their organizational structure as I did in the London Unit.

5.2  Stay at the Nuclear Studies Center in Saclay, France

In Saclay, the Nuclear City of France there was an "Arithmetical Electronic Service" where I spent the second part of my fellowship.

Programming and technical work The machines in Saclay were: IBM’s 1401, 1620, 704 and 7090 and one Ferranti Mercury. The work system was "closed store", so I could never visit the Computing Centre, nor run my programs in any of their computers.. In view of the equipment they had, a short period of my stay in Saclay was devoted to learning and practicing FORTRAN, a high level language that had been in use in the computing world since 1959. The subject in Numerical Analysis the Senior Analyst of the Mathematical staff, with whom I was assigned to work suggested was the study and adjustment of a method in Approximation of Functions. This part of the fellowship was completed at the Blaise Pascal Institute, were the Francis method for solving matrix eigenvalues was translated to Fortran, tested and used by the Institute.

Courses and lectures Two complete courses in Numerical Analysis were attended to during that second part of my fellowship. They took place in the Henri Poincaré Institute, and they were: a) Approximation methods to solve Partial Differential Equations of the elliptic type, given by Lions, a famous French researcher.. b) Matrix Calculus, taught by Rigal.

6  End of Clementina´s era

Political changes in our underdeveloped countries always bring important changes in authorities, in educational institutions and in the Universities. A political event also marked the end of Clementina, but The Mercury continued operating in other places of the world. Our Mercury was beginning to be dismantled shortly after the Institute of Calculus had an important change of staff following a military coup.

7  Epilogue

In 1966, a military coup and the brutal invasion of the army of some institutions, one of which was the School of Natural and Exact Sciences and our Institute of Calculus, were students and professors were beaten and injured, had disastrous effects. In our Institute, the 90% of the scientists, professors and staff resigned and took on relevant positions in the country and abroad, working in private and state institutions. Whole teams of great scientific importance emigrated and went to enrich other communities. Most of them left the country and took with them their knowledge, expertise and geniality. Those groups were successful in the country and abroad in everything they undertook. Our mentor, Manuel Sadosky, was exiled, first in Venezuela and afterward in Barcelona, Spain. In Barcelona he took part in the creation of a Science Museum for Children. In Venezuela a group of scientist from our Institute filled important positions at the Simon Bolivar University and the Central University of Venezuela; where they had successful careers.

I stayed in the country where I began a successful managerial and counselling career, but I always kept in touch with my working team of the University wherever they were exiled, specially with my teacher and friend Manuel Sadosky, who was invited later in Uruguay, where he repeated the experience of the Institute of Calculus, joined the University of the Republic of Uruguay and obtained an "Honoris Causa" title in Montevideo.  

References

Nicolás Babini. La Argentina y la computadora. Editorial Dunken, New York, NY, USA, 2003.

Nicolás Babini. La llegada de la computadora a la argentina, volume 2. LLULL, New York, NY, USA, 1997.

Berg E Kerr R.H. Broker R.A., Richards B. Mercury autocode manual.

W. O. Durand. Introducción al lenguaje comic. 14.

Sadosky M. Cinco años del instituto de cálculo de la universidad de buenos aires. 14, 1972.

V. Pereyra C. Berdichevsky G. Oliver E. Ruspini G. Galimberti Zadunaisky, P. Un método para la estimación de errores propagados en la solución numérica de un sistema de ecuaciones ordinarias. 1, 1964.