Scientific activity of V.M. Glushkov Glushkov Viktor Mikhailovich

A Universal Small Electronic Calculating Machine (MESM) and a Specialized Electronic Calculating Machine for solving algebraic equations, the first matrix-vector processor in the USSR with a conveyor organization of calculations and a combination of data entry and calculations (SESM) were created according to the ideas of academician S.A. Lebedev, the first - under his direct supervision, the second - completed under under the leadership of academician B.V.Gnedenko, V.M. Glushkov took part in the formation of the results of these works and was the responsible editor of the book initiated by him "Specialized electronic counting machine SES" / Z.L.Rabinovich, Yu.V.Blagoveshchensky, R.Ya.Chernyak, A.L.Gladysh, I.T.Parkhomenko, I.P.Okulova, L.A.Mayboroda, S.S.Zabara. - Ed. Academy of Sciences of the Ukrainian SSR Kiev. 1961. -147s.

This book was republished in English in the USA and was one of the very first Soviet books on computing published abroad.

A Specialized Electronic Calculating Machine (SESM)

Computer "Kiev"

From the memoirs of V.M. Glushkov

... The Kiev computer played a significant role in the development of the Computer center's work, although it did not go into mass production. The Institute first entered the All-Union market with this computer, the second copy of the machine was bought by the International Institute of Atomic Research in Dubna. In 1956-1957, atomic physics was "booming", so working with this institute helped a lot and taught a lot. On the one hand, the Institute of Cybernetics was doing high science, and on the other, it was learning to work with industry.

In 1959, at the Computing Center of the Academy of Sciences of USSR, work was completed on the creation of the first Kiev mainframe computer in Ukraine.

It was on the Kiev computer, in addition to the effective solution of computational problems, that the first experiments on automated design of electronic circuits were carried out, tasks on recognition of visual images were solved, the first database "avtodirektor" operated, the experience of controlling the habits of the Bessemer converter in Dneprodzerzhinsk (for the first time in Europe) and the control of the soda carbonation process was carried out in Slavyansk. The customer of the second copy of the Kiev machine was the well-known Joint Institute for Nuclear Research in Dubna...

Computer "Kyiv"

Computer “Kyiv” - an electronic digital computer designed to solve a wide range of scientific and engineering problems. Developed in 1958. at the Institute of Cybernetics of the Academy of Sciences of the Ukrainian SSR on the initiative and under the guidance of B.V. Gnedenko, chief designer - L.N.Dashevsky. The development was initially carried out by the same team that created the MESM; V.S. Korolyuk, I.B. Pogrebysky, E.L. Yushchenko participated in the selection of operations — employees of the Institute of Mathematics of the Academy of Sciences of Ukraine. V.M. Glushkov joined at the final stage of technical design, assembly and adjustment of the machine and took an active part, being, together with L.N. p>

The machine was intended for the organized Computing Center, used for the first time in the USSR for research on remote control of technological processes.

Computer “Kyiv” represented a significant new word in computing — had asynchronous control, ferriteyu RAM, external memory on magnetic drums, input — decimal number output, passive memory with a set of constants and subroutines of elementary functions, a developed system of operations, including group operations with address modifications performed on complex data structures, etc.

Characteristics of the computer "Kyiv"

• Parallel RAM based on ferrite cores with a capacity of 1024 words, RAM access cycle - 10 microseconds.
• Reduced multiplication and division operations have been implemented.
• The command structure is three-address.
• To “Kyiv” For the first time, an address programming language was used as an input language for a translator. The system of machine operations - 32 operations, including access to the address of the 2nd rank and operations for setting cycles.
• The form of representation of numbers is with a fixed comma before the most significant digit, the length of the machine word is constant, 41 binary digits.
• Floating point mode is done by software.
• A permanent (single-sided) memory of a ferrite-transformer type with a circulation cycle of 7 microseconds and a capacity of 512 words is designed to store replaceable-soldered programs.
• The cycle of the machine is four-stroke, the duration of the cycle is variable, it depends on the type of operation and the memory used.
• The parallel arithmetic unit includes a push-pull accumulator and 3 registers; addition time 6.6 microseconds, division time - 275 microseconds, average speed 15 thousand operations per second.
• The external memory consisted of three magnetic drums with a total capacity of 9 thousand words with an access time of 120 microseconds.
• Element base - impulse-potential (tube elements).
• Data input is carried out from punched tapes, punched cards, telegraph communication lines, continuous-to-discrete converters, graph readers.
• Output device - digital printer or punch.

Computer "Kyiv" became the first system in Europe for digital image processing and modeling of intellectual processes. 

On the computer "Kyiv" under the direction of V.M. Glushkov in the late 50s - early 60s, a series of works on artificial intelligence was carried out, in particular, learning to recognize simple geometric shapes (V.M. Glushkov, V.A. Kovalevsky, V.I. Rybak), modeling readers automata for handwritten and typewritten characters (V.A. Kovalevsky, A.G. Semenovsky, V.K. Eliseev), tracking the movement of objects on a series of images, or cinematography (V.I. Rybak), modeling the behavior of a group of automata in the process of evolution (A.A. Dorodnitsyna, A.A. Letichevsky), automatic synthesis of functional circuits of computers (Yu.V. Kapitonova), the first database of the relational type "Autodirector" (V.G. Bodnarchuk, T.A. Grinchenko), etc. .

The principles of construction of the Dnepr computer, its main parameters, structure and architecture were determined by its purpose - management of a wide range of production processes.

The main ideas that formed the basis for the development of the Dnepr computer are as follows:

• the computer must be semiconductor,
• transportable,
• with highly reliable protection,
• low-bit (26-bit - then it was enough to control the technology in most processes).

The main feature of the computer was the presence of a universal communication device with the object - USO - a set of analog-to-digital and digital-to-analog converters controlled from the machine, with which the machine was connected to the production process.

From left to right: V.I.Skurikhin, L.A.Korytnaya, L.A.Zhuk, V.S.Kalenchuk, V.M.Glushkov, B.N.Malinovsky

During the development of the USO, the need for standardization of electrical signals at the output of measuring instruments and at the input of serving mechanisms acting on the process immediately became obvious. Only in this case, the creation of an RCD designed to receive many input and output many output signals became possible.

In the process of developing the computing part (arithmetic device and memory) the computer created a ferritic memory on miniature cores. The memory device on miniature ferrite cores was the first in the country and provided high reliability and small size of the machine.

For the input-output device of programs and data, a data input device with punched tape and digital printing with an electric drive were widely used at that time.

Characteristics of the computer "Dnepr"

Date of creation:1961

Purpose: The digital semiconductor control computer of a wide purpose "Dnepr" was developed at the Institute of Cybernetics of the Academy of Sciences of the Ukrainian SSR. Designed for monitoring and controlling continuous technological processes and complex physical experiments, as well as for studying processes during their algorithmization

• Command structure: two-address
• Number system: binary
• Method of representation of numbers: with a comma fixed before the highest digit
• Bit depth: 26 binary digits, of which one is the highest -signed
• Performance: when controlling (on/ off) - 50,000 operations / sec; when adding and subtracting - 20,000 operations /ec; when multiplying and dividing - 4000 operations/ sec. Average speed of 10000 operations/sec.
• Command system: 88 commands
• Input devices: from telegraph punched tape (17.5 mm) at a speed of 45 digits/sec, from the keyboard of the telegraph machine and from communication lines
• Types of elements used in the computer: pulse-potential
• Occupied area: 35-40 sq. m. meters

• Power consumption: 4 kW

Creators of the computer "Dnepr" (UMSHN):

Certificate of registration N 30882

Issued by the Committee for Inventions and Discoveries under the USSR Council of Ministers on the recommendation of the Institute of Cybernetics of the Academy of Sciences of the USSR.

Work under the name "Wide-purpose control machine UMSHN"

Project managers:

Glushkov Victor Mikhailovich (scientific supervisor), Malinovsky Boris Nikolaevich (chief designer).

Project team:

G.A.Mikhailov, N.N.Pavlov, B.B.Timofeev, A.G.Kukharchuk, E.S.Oreshkin, V.S.Kalenchuk, L.A.Korytnaya, V.M. Egypko, L.A. Zhuk, S.S.Zabara, L.Ya.Pripada, E.P.Raichev, N.M.Abakumova, L.A.Rusanova, G.I.Kornienko, F.N.Zykov, V.S.Lenchuk, I.D.Voitovich, V.V.Kraynitsky, A.A.Pushchalo, Yu.T.Mitulinsky, E.P.Dragaev, A.I.Tolstun, M.A.Ermolenko, N.K.Babenko, E.F.Kolotushchenko.

From the memoirs of V.M. Glushkov

... Simultaneously with theoretical research at the Institute of Cybernetics of the Academy of Sciences of Ukraine, work was launched on the creation and application of computer technology in Ukraine. At that time, the simplest analog computing devices were used to automate the control of technological processes. A special device was created for each process. And mainly for those described by differential equations (not very complex).

Therefore, when I put forward the idea of creating a universal control computer (UMSHN) at the All-Union Conference in Kiev in 1958, it was met with hostility. Many specialists in the field of computer technology unanimously opposed it. The fact is that at that time the universal machine was necessarily represented by a lamp, and this required huge halls, air-conditioned air, i.e. it was not linked in any way with production and process control.

The development of the machine was entrusted to B.N. Malinovsky, he was the chief designer, and I was the scientific supervisor. The work was completed in record time: from the moment the idea was expressed at a conference in June 1958 to the moment the computer was put into production in July 1961 and installed in a number of productions, only three years passed. As far as I know, this result remains a world record for the speed of development and implementation until now. In parallel with the creation of the UMSHN, which later received the name "Dnepr", a large preparatory work was carried out with the participation of a number of Ukrainian enterprises on its application for the management of complex technological processes. As an experiment, for the first time in Europe, on my initiative, remote control of the Bessemer process was carried out for several days in a row in the mode of the master's adviser. Research has begun on the use of Dnepr machines for the automation of plaza work at the Mykolaiv plant named after 61 communards. B.N. Malinovsky, V.I. Skurikhin, G.A. Spynu and others participated in them.

"Dnepr" became the first Soviet semiconductor machine (except for special machines), it perfectly withstood various climatic conditions, shaking, etc. Then it turned out that the Americans had started work on a universal semiconductor control machine similar to the “Dnepr” a little earlier than us, but they put it into production in June 1961, at the same time as us. So this was one of the moments when we managed to reduce to zero the gap that exists in relation to American technology, albeit in one, but very important direction. Note also that our computer was the first domestic semiconductor computer (except for special computers).

This first universal semiconductor computer, which went into production, also broke another record - the record of industrial longevity, since it was produced for ten years (1961-1971), whereas this period usually does not exceed five or six, after which serious modernization is required. And when, during the joint space flight "Apollo-Soyuz", it was necessary to put in order the showroom in the Mission Control Center, then after a long selection of machines that existed at that time (in 1971 or 1972 this work began) the choice still stopped at the “Dnepr”, and two cars controlled a large screen on which everything was displayed - docking, etc. (the system was made under the guidance of A.A. Morozov). This computer was exported and worked in many socialist countries.

The first computers "Dnepr" were produced by the Kiev plant "Radiopribor". Simultaneously with the development of the Dnepr machine, a plant of computing and control machines (VUM) - "Electronmash" began to be built in Kiev, on the initiative of the Institute of Cybernetics, supported by the government. So the development of "Dnepr" marked the beginning of a large computer manufacturing plant...

The history of computer technology does not even count sixty years. But there are so many events in this very short interval that the time separated from us by four decades seems almost legendary. It was then that the world's first personal computer was created in Kiev at the school headed by Victor Mikhailovich Glushkov. However, it was distinguished from the now familiar desktop computer by its much larger size. The finest hour of microelectronics had not yet come, and the potential semiconductor elements on which the MIR computer (a Machine for Engineering Calculations) was built determined its size.

But the most amazing thing was still the idea of the creators of this computer. The high level of the internal language, the developed means of dialogue with the user, the developed operating system, the effective system of microprogramming provided the computer MIR with high machine intelligence and were far ahead of the then level of world computing technology. Since 1969, when the MIR-2 computers and its subsequent modifications began to be mass-produced, users were able to solve their tasks in the mode of direct interaction with the computer.

Computer MIR-1.

Computer MIR-2.

Computer MIR-3.

The developers of the MIR computers solved rather difficult and unusual tasks of equipping these machines with the necessary software tools for their work, some of which were transferred to the level of hardware and software implementation during the creation of computers with the help of microprogramming. At this crucial stage, the external and internal languages were joined. The effectiveness of the docking determined the effectiveness of the plan. And it was here that new ideas were born, which became the basis of the ANALYTIC language focused on the automation of analytical transformations.

ANALYTIC, apparently, was the first programming language that dealt not with linearly ordered records of algorithms, but with texts in which the linear order is not rigid. That is why there are opportunities in it that did not find their development in the 60s, because the time has not yet come for them. The idea that most texts are network formations, hypertexts and capable of generating many different linearly ordered texts had not yet arisen. But now that hypertext technology has taken an important place in the practice of solving problems on computers, such capabilities of the ANALYTIC language are becoming very transparent.

The history of the MIR computers is a unique event, one of the most remarkable pages in the history of computers. V.M. Glushkov's outstanding contribution to the creation of the MIR series of computers is determined not only by the fact that he formulated and theoretically justified concepts that were ahead of their time, but also by the atmosphere of creative uplift created by him, which contributed to the effective implementation of the project and left an indelible mark in the memory of all developers.

From the memoirs of V.M. Glushkov

... In 1959, I had a program of work on machines for engineering calculations. It was started with the development of a digital computing machine. And in 1963, we launched the Promin computer into mass production.

When it was ready, the Severodonetsk Plant of Computers  began to produce it. It was essentially a new word in world practice, had a number of technical innovations, in particular memory on metallized cards. But most importantly: it was the first widely used machine with the so-called step-by-step firmware control (for which I later received an author's certificate).

Unfortunately, we did not patent the new control scheme, since we were not a member of the International Patent Union at that time and could not patent and acquire licenses. Later, step-by-step firmware control was used in the MIR-1 computer for engineering calculations, created after the Promin computer (in 1965).

In 1967, at an exhibition in London, where the MIR-1 machine was demonstrated, it was bought by the American firm IBM, the largest in the United States, which is a supplier of almost 80% of computing equipment for the entire capitalist world. This was the first (and, unfortunately, the last) purchase of a Soviet computer by an American campaign.

In 1969, a new, more advanced MIR-2 computer was put into production. Then the MIR-3 computer was developed. They had no competitors in terms of the speed of performing analytical transformations. The MIR-2 computer, for example, successfully competed with mainframes of the usual structure, surpassing it in nominal speed and memory capacity by hundreds of times. For the first time in the practice of domestic mathematical engineering, an interactive mode of operation using a display with a light pen was implemented on this computer.

COMPUTER MIR-2.  I/O device with a light pen (prototype of a modern touchscreen)

In small computers of the MIR series (MIR-1, MIR-2, MIR-3), it was possible to fully implement the hardware implementation of high-level languages. The structural interpretation of the high-level languages MIR and ANALYTIC allowed us to obtain an effective implementation of working with real numbers of arbitrary bit depth, integers of unlimited bit depth and precise operations on fractional rational numbers, etc. In the MIR-2 computer, operations were performed not only on numbers, but also on arbitrary algebraic expressions, which were considered up to the basic relations of the algebra of analysis (including relations for transcendental functions). The structural implementation of analytical transformations provided such a significant increase in productivity that for a number of specific analytical tasks, their solution time on the MIR-2 computer turned out to be comparable to the solution time on traditional architecture computers with nominal performance hundreds of times higher than the MIR-2 performance.

Each of these computers was a step forward in the direction of building a reasonable computer - our strategic direction in the development of computers.

In the creation of MIR computers , the remarkable foresight of V.M. Glushkov was manifested, who personally drew up an advance project of the first model of MIR, in which he outlined the basic principles of building such computers as machines of a new type with the implementation of high-level language in them in a structural way. The principle of structural interpretation of high-level language was proposed by V.M.Glushkov and the participants of his school in 1962 (even before the creation of the MIR computers) and was first published in 1965.

It is characteristic that the first proposal, issued in the form of an author's application (and then a closed certificate ), referred to the implementation of high level language in a high-performance (i.e., large in the terminology of the time) computer with a high degree of versatility. Our proposals were widely discussed by the scientific community as "revolutionary". Some doubts were expressed about the possibility of creating such a high-performance machine at that time. Such doubts were also expressed in a foreign forecast based on the insufficiency of the element base at that time.

V.M. Glushkov decided, as it turned out later, the only correct solution to this scientific and strategic problem: to force the creation of a small computer MIR in every possible way (as computers for mass use) and at the same time to carry out a detailed development of a universal high-performance computer with an internal language similar to ALGOL-60, expanded by including array processing tools in it, strings and words of variable bit depth, as well as computational process controls. The choice of ALGOL-60 as the main input language of the computer was due to its advanced level at that time, universalism and widespread (in order to be able to effectively use the accumulated user support).

This computer was named "Ukraine", but it was implemented only at the level of a technical project, supported by layout and modeling. The project contained the embodiment, along with the principle of direct structural interpretation of high level language, of a number of other related advanced ideas - a virtual memory field dynamically mapped into multi-stage physical memory, fully automatic addressing, a widely developed set of operations with parallel and sequential execution, combining management and information processing processes, etc.

The project was subjected to a tough discussion in the Ministry of Radio Industry, was fully approved and sent to a number of leading organizations for possible use. Thus, the idea of implementing high level language not only in small, but also in high-performance computers has become generally recognized and has been reflected in several advanced domestic developments.

Unfortunately, the computer "Ukraine" was not built at the Institute of Cybernetics and was not put into mass production, due to the following reasons:

- the lack of a sufficiently miniature and high-speed hardware at that time;

- difficulties in financing.

The latter circumstance was aggravated by the fact that the country's need for high-performance computers was already largely satisfied due to the launch of BESM-6 serial computers developed at the Lebedev Institute and located at the most advanced edge of the world computing technology. And although this computer did not reach the level of intelligence envisaged in "Ukraine", it had a very effective architecture, including a developed internal language.

Therefore, V.M. Glushkov, in the light of the state approach peculiar to him, did not insist on the expensive creation of the "Ukraine" machine, which would differ favorably from the BESM-6 only with increased intelligence due to the structural interpretation of the input high level language. By this time, the problem of intelligence, especially acute for the mass use of computers by specialists of various specialties, was already largely solved by the serial production of the MIR computers.

The development of the computer "Ukraine" was an important milestone in the development of the scientific school of V.M. Glushkov in the field of computer technology. The ideas laid down in the project anticipated many of the ideas used in American computers of the 70s. Based on the materials of the development, a monograph "A computer with advanced interpretation systems" was prepared, the authors of which are V.M.Glushkov, A.A.Barabanov, S.D.Kalinichenko, S.D.Mikhnovsky, Z.L.Rabinovich, published in 1970.

From the memoirs of V.M.Glushkov

... The basis of our further work on computer architecture, I put a consistent rejection of the well-known von Neumann principles (sequential structure of the language, i.e. execution of commands one after another; command-address principle, i.e. the command contains the addresses of operands, and commands are stored in the same way as operands in memory; maximum simplicity of the system commands, i.e. maximum simplicity of the computer language. We can talk about other principles, but these are the main ones). The appearance of such principles is not surprising. In the era of lamp machines, when each digit of an arithmetic device is at least one triode, a simple machine with simple commands was needed.

However, even then I foresaw the development of microelectronics and the fact that the structural elements would be manufactured in a single technological process and would cost very cheap. Back then, I formulated such a goal for physicists: compositional construction of a solid body to create a computer environment. In this case, von Neumann's principles are unacceptable. As one of the new principles, I proposed a complicated computer language, because compiling systems were becoming more complicated, and it was necessary to simplify programming from both ends, from the point of view of languages and compilers, i.e. to bring the computer language closer to the input. Having partially implemented this idea in the MIR series of computers, we began to develop it further in accordance with the principle of gradual complication of the computer language, and not just complication, but approximation to the human language. The limit I set was a conversation with the machine in natural language (and the issuance of tasks)...

The second direction is related to the dynamic parallelization of sequential programs. The idea proposed by V.M.Glushkov is to use structured programs for parallelization, presented as expressions of the algebra of algorithms.

These ideas have found their implementation in high-performance multiprocessor computers with macroconveyor computing organization. Although five years ago the need for the development of supercomputers was questioned, at present it is already clear that the world computer building in the field of high-performance computers has followed this path.

In terms of performance, the MACROCONVEYOR surpassed Elbrus (S.A. Lebedev, ITM and VT).

Unfortunately, despite the production of experimental and serial samples, the complex was not put into mass production due to the beginning of "perestroika" (which also interrupted the completion and launch of other promising developments of the Institute of Cybernetics, carried out at the state level).

MACROCONVEYOR project:

Supervisors: V.M. Glushkov, V.S. Mikhalevich

Chief Designers: S.B. Pogrebinsky, A.G. Kukharchuk

Deputy Chief Designers: V.D. Losev, V.P. Klimenko

Heads of directions: Yu.V. Kapitonova , A.A. Letichevsky, I.N. Molchanov

Developers: Reutov G.V., Alexandrov V.Ya., Orlova I.A., Pereloma A.A., Burachenko T.E., Borsch N.S., Mayboroda V.T., Gritskov V.Ya., Shmidsky Ya.K., Dorodnitsyna A.A., Reznik A.M., Kalnenko V.P.

From the memoirs of V.M.Glushkov

... Automating the proof of theorems is my blue dream, it forms the basis in my thoughts about the architecture of new computers capable of carrying out complex creative processes, including the construction of deductive theories.

It is from here that new ideas of computer construction follow. And only a person who is engaged not only in computer, but also in artificial intelligence can understand how to build such computer. This is our strength.

In addition to complicating the machine language, we tried to move from the sequential command execution principle proposed by Neumann to the multi-command one. It took a lot of work until the idea of a macroconveyor came to mind, and it was possible, if not for each arithmetic device, then for the whole system as a whole to make a multi-command computer with many streams of commands and data.

The essence of the principle of macroconveyor data processing proposed by me is that each individual processor is given a task at the next step of calculations that allows it to work autonomously for a long time without interacting with other processors...

The PROJECT-1 system, implemented on the M-220 and BESM-6, was a distributed specialized software and hardware complex with its own operating system and a specialized programming system. In it, for the first time in the world, the algorithmic design stage was automated (and with optimization). Within the framework of these systems, a new technology for designing complex programs was developed - the method of formalized technical tasks. The "PROJECT" systems were developed as experimental, real methods and methods of designing circuit and software components of computers were worked out on them. These methods and techniques were subsequently adopted in dozens of organizations developing computer technology. The customer was the Ministry of Radio Industry (CDB "Almaz" and NITSEVT). The developed systems became the prototype of real technological lines for the production of documentation for the production of computer chips in many organizations of the former Soviet Union.

The PROJECT-1 system is closely connected with the automation system for the design and manufacture of BIS using elion technology. In the department headed by V. P. Derkach (one of the first graduate students of V.M. Glushkov), the Kiev-67 and Kiev-70 systems were created that control the ion beam when processing various types of substrates with it. It should be noted that the indicators of these systems gave record parameters in microelectronics at that time. The PROJECT's design automation systems had a communication interface with Kiev-67 and Kiev-70, which made it possible to perform complex programs for controlling the ion beam, both during spraying and during graphic processing of substrates.

The main features of the development and implementation of computer design automation are the use of algebra calculations, discrete converters and other information processing tools. The system approach implemented in computer-aided design systems implies integrated consideration and presentation in the system of both objects and design operations at various stages of the design process with the necessary maintenance of the process itself. Moreover, these solutions should reflect not only existing technologies, but also their development in the future.

The creation of the information processing industry required a qualitatively new level of unification and typification of solutions for technical, mathematical, information and organizational support of systems, and this is primarily due to the use of mathematical apparatus in the design process.

Development managers: Kapitonova Yu.In, Letichevsky A.A.

The main developers of the PROJECT system: Grebnev V.A., Chebotarev A.N., Gorokhovsky S.S., Mishchenko N.M., Gorlach S.P., Kolbasin N.I., Chuikevich V.S., Godlevsky A.B., Lyabakh V.F., Mitchenko A.I., Mishchenko A.T., Morozov S.I., Parnitsky V.I. Pyatygin S.A., Rystsov I.K., Berestovaya S.N., Valkevich T.A., Doroshenko A.E., Krivoy S.L., Krat S.P.

FPGA-based reconfigurable systems are widely used in many areas: reconfigurable data processing; digital signal processing; image processing; communications; general purpose computing devices; verification.

FPGA is a programmable logic integrated circuit that combines the regularity of the structure of a semiconductor memory device with the versatility of a microprocessor, which allows you to programmatically form an internal specialized processor. Structurally, FPGAs are a homogeneous environment and have the following properties: homogeneity, reconfigurability and parallelism of operations. Parallelism - increase in speed, is achieved by increasing the clock frequency and by parallel execution of a large number of operations. Reconfigurability - reliability, flexibility and structural versatility (the ability to create an appropriate structure for each task) are provided in hardware by changing the links between the elements and the functions of the elements themselves. Homogeneity - the simplicity of manufacturing technology when using the same elements and the same type of connections between them.

Today, FPGAs have gone from simple PLA elements (programmable logic arrays) to complex FPGAs - CPLD (Complex Programmable Logic Devices) and FPGA (Field Programmable Gate Array). Xilinx's Virtex-II-Pro dies are available with capacities up to 10 million system gates, which have a huge advantage in logical capacity over previous FPGA dies. Within the next three to four years, devices with a capacity of 50 million system gates will be offered - enough logic to form complex, high-performance systems on a single chip.

The essential advantage of FPGAs is their versatility and the ability to quickly program (configure) for a given project. The availability of design automation tools that are extremely convenient for the developer allows you to quickly and efficiently develop, verify and debug a project at one workplace, using a PC or a workstation as the main technical tool.

The possibility of multiple reprogramming (configuration) allows you to make changes to an already finished, functioning product or use this product to perform various functions depending on the loaded project.

The Virtex die family (including the Virtex, Virtex-E, Virtex-EM, Virtex-II, Virtex-II-Pro series) is optimized for both hard and soft cores. Hard cores (such as PowerPC) are specialized areas of the FPGA chip dedicated to certain functions, in which blocks of a fixed structure are created that are optimized for a given function.

The accumulated experience has allowed FPGA manufacturers, relying on the experience of hardware developers, to recommend FPGAs to replace microcircuits of a small and medium degree of integration in the implementation of standard combinational ones (encoders, decoders, adders, code converters, comparators and etc.) and sequential (registers, shift registers, counters, etc.) functional nodes, as well as glue logic. For this purpose, CORE Generator is introduced into the FPGA CAD system - a tool that provides the user with parametric logical "kernels" optimized for FPGA crystals.

It should be noted that the CORE system, thanks to the use of network technologies, can significantly reduce the development time for new projects. In accordance with the formulated technical requirements, a designer can obtain an FPGA-optimized logic core over the Internet and include it in his project.

When designing digital devices based on FPGAs of varying complexity - from simple PLDs to CPLDs and FPGAs, hardware description languages (HDLs) are widely used. The most popular HDL languages are VHDL, Verilog and Abel.

Reconfigurable Computing. The rapid development of modern technologies and the production of highly integrated FPGAs has led to the creation of new directions in "Computer Science" - "Reconfigurable Computing" and "IRL - Technology". The term "Reconfigurable Computing" generally refers to the concept of both a reconfigurable computer structure (hardware) and the data processing process performed by the computer. The IRL (Internet Reconfigurable Logic) technology provides for the possibility of reconfiguration (including remote, via the Internet) of the structures of computing devices included in this network and implemented on the element base of an FPGA type FPGA.

Theoretical foundations and examples of practical implementation of computers with a flexible (programmable) architecture, which uses the mechanism of microprogram emulation as a tool for restructuring the architecture, primarily the processor, were developed under the guidance of A.V. Palagina. This provides tuning to various internal languages, compatible in terms of bit depth, popular at that time mini-computer models (DEC, H / P, etc.), i.e. the result was a kind of "polyglot machine" that could "digest" software implemented in various internal languages (command systems), creating the problem-oriented configurations necessary for each specific application. Another application of the flexibility property was the development of the internal language of the basic models of one of the well-known families of original microcomputers as an effective means of increasing their performance.

The main component of the architecture of this class of computers was a flexible hierarchical control system using the original "emulator" of command formats at the lower, microprogram level of control.
The theoretical basis for the development and evaluation of the effectiveness of a computer control system was the proposed logic -informational design method. In accordance with the logical concept of the logical-information model, the processor is represented by a well-known composition of the operational and control automaton. In accordance with the information concept, the processor is considered as an information system, all information in which is assigned to three "spheres" of states: storage, transportation, and transformation. Thus, with certain relationships between information objects in these areas, it is possible to obtain the optimal technical parameters of a computer.

An important quality of a computer with a flexible architecture is the degree of flexibility, or the level of programmable components. It is they that determine the range of technical solutions and properties of architectures, each of which is effective in its own, quite specific, class of problems. In this case, the level of programmable components descended to the functional units of the computer control system, so it can be conditionally called the "automatic" level. The basic components at this level in the 70s were PLA programmable logic arrays.

With the advent of modern FPGA chips with a capacity of more than 100,000 logic gates, it became possible to use the results obtained to build reconfigurable computers with a fully programmable architecture. In reconfigurable computers (computers with a programmable architecture), a processing field of a given dimension is fixed, configured specifically to execute a certain given algorithm or part of it, thus ensuring that this algorithm is implemented in an optimal way, bearing in mind both its execution time and the cost of hardware resources. The algorithm can be divided into fragments that are executed sequentially, and therefore, the structures corresponding to these fragments are also loaded into the crystal sequentially (in the order of their execution), which leads to significant resource savings. The complexity of the fragments of the algorithm is determined only by the logical capacity of the crystal, i.e. processing field dimension.

Thus, reconfigurable data processing is, to a certain extent, a change in the central design paradigm of modern computer technology and automation, and reconfigurable hardware is becoming a real and rapidly developing area of ​​computer technology. The process of configuring FPGAs, which form the basis of reconfigurable devices, can be implemented if there are appropriate configuration files obtained in the process of creating a project using CAD.

The differences between programmable microprocessors and FPGAs are gradually blurring. Modern FPGA chips include increased local memory, specialized multipliers and RISC processors (Power PC). Most likely, FPGA crystals will neverThey will not replace microprocessors for general purpose computing, but the concept of configurable data processing is likely to play an increasing role in the development of high-speed reconfigurable computing systems. Similar to Internet-connected computers that can automatically download software components to perform specific tasks, reconfigurable devices can download new hardware configurations as needed using IRL technology.

Principles of building reconfigurable systems

Currently under the direction of A.V. Palagin is working on creating a computer system with a reconfigurable (virtual) architecture, which is a problem - oriented configuration in relation to each specific task. The structure of a reconfigurable system consists of 2 parts: a permanent (or "fixed") part F - the host of the computer and a variable part V - the so-called "reconfigurable" equipment, which can be combined into various configurations. Equipment V is also divided into two parts: the "standard" part, which is connected to F via the computer's standard host buses, and is represented by the motherboard with a local internal bus for connecting the "non-standard" part, which is a wide range of expansion modules. The operations performed in each of the parts are determined by the following characteristics: in F, the computation time and initial data; in V - also additional equipment necessary to perform the corresponding operations, the time of information transfer between computing modules and the time of system reconfiguration (loading soft cores into FPGA crystals).

In this system, the configuration is formed in such a way as to transfer the main work from the F - part of the system to specialized blocks (V - part), which are soft cores. For a strictly formulated computational problem (where all numerical procedures are uniquely defined) and a description of the characteristics of operations for F and V, it is required to organize a common structure ( ) and distribute calculations in such a way as to minimize the objective function (the sum of reconfiguration costs and computation time).

This problem is extremely complex, in essence it is a combinatorial problem of optimal synthesis. The limitation imposed by the finite volume of reconfigurable equipment does not allow one to obtain an unambiguous method acceptable for practice for finding the optimal solution. Therefore, a solution (close to optimal) is found by the method of successive approximations.

Structure reconfiguration has two phases. In the first one, only a part of changes, i.e. mechanical modification is not allowed. If the specified optimization criterion is not achieved, that is, the part does not have sufficient logical power and memory, or specific means of input-output of information, then a transition to the second phase is carried out. And further reconfiguration of the system is also performed mechanically (by installing expansion modules in the appropriate slots - the local bus slots of the motherboard).

Structural organization of reconfigurable processors. Reconfigurable processors (RP) are in the minimum configuration a printed circuit board with one or more custom FPGA chips (FPGAs), non-volatile memory for storing configuration files, elements for loading the configuration file (files) and one or more connectors for connecting external devices ( expansion modules). The type of non-volatile memory is determined by the scope of the RP: for dynamic configuration of the FPGA during operation, it is advisable to use Flash - memory, and in the absence of such a need - EPROM. The use of Flash-memory implies the presence of a control unit for this memory in the RP, which implements the loading of this memory with configuration files from an external source, as well as reading with random access of the required file and loading it.

Most data processing tasks require cache-memory as part of the RP. The memory must be accessed both from an external device (through the installed connector) and from the device implemented in the FPGA. In order to expand the memory, a connector for connecting additional memory is installed on the RP board. Cache - memory as part of the RP assumes the presence of a memory controller.
This type of RP is called carrier boards or motherboards, due to the fact that expansion boards can be connected to them. RPs are designed for industry standards such as PCI, CompactPCI, PMC (PCI Mezzanine Cards), DIME (DSP and Image processing Module for Enhanced FPGAs) and VME. The PMC specification allows expansion modules to be added to motherboards via the local PCI bus. Motherboards are connected to the standcomputer bus and work in coprocessor mode.

The set of main tasks in this area is formulated as follows:

- development of theoretical foundations for the principles of building reconfigurable digital structures based on a homogeneous environment, in accordance with this, build a system of formalized methods and algorithms for the synthesis of parametric modules, taking into account the features of their constructive and technological base;
development of the foundations the theory of adaptive logical networks (ALN) designed to solve a wide class of problems by direct structural implementation of algorithms for processing and mapping the input data set to the output data set due to their structural versatility, the structural organization of which is based on the requirements of dynamic reconfigurability, multilevelness and parallelism of data processing , which fully corresponds to the modern element base - FPGA;
- development of algorithms for adapting logical networks to given classes of tasks, including classification tasks, etc.;
- development of the structure of a reconfigurable processor, realizing its pipeline principle of data processing;
- development of basic library parametric modules by describing them in the VHDL language and the FPGA CAD circuit editor, including a threshold device, a Hamming adder, sorting devices, median filters, matrix multipliers and others;
- development of the fundamentals of the structural organization of neuro-like Hamming networks based on FPGA crystals.

The main differences between the results of work in this area are:

- orientation to the technological capabilities of domestic enterprises of computer engineering;
- the property of structural universality (in addition to algorithmic universality), which allows for each algorithm to create its own structural diagram in the universal functional field, which provides an equivalent display of the structural schemes of the algorithm and functioning processes, allowing you to change the logical structure of the device depending on the specifics of the problem being solved by reconfiguring the internal structure.

Reconfigurable systems will find wide use in the following applications:

- problem-oriented systems and coprocessors;
- telecommunications, image processing, digital signal processing;
- modeling algorithms and designing architectures of modern computers based on FPGA and ASIC chips;
- modern control systems associated with the performance of large amounts of calculations - control of technological processes, instrumentation, robots - manipulators, other real-time systems;
- knowledge - oriented systems - creation of automatic networks for syntactically - semantic analysis of texts, in particular, for the implementation of the model of the language picture of the world (LCM).

This direction is carried out within the framework of works that are a theoretical generalization of scientific and practical results obtained over more than a ten-year period of development of a number of digital devices based on FPGAs, carried out in the department of "Microprocessor Technology" of the Institute of Cybernetics. V.M. Glushkov National Academy of Sciences of Ukraine.

Then we began to think over the Lviv system (ASU) with a large-scale nature of production at a television factory in Lviv.

In 1965, I put forward the concept of a specialized operating system designed for systems with a regular flow of tasks plus a small percentage of irregular tasks. The fact is that the operating systems with which IBM-360 computers were supplied in 1965 and which solve random streams of tasks are universal for batch mode and are good for computing centers (relatively good, of course). And in the automated management information system, as a rule, we dealt with regular tasks, i.e. we knew that at some time such and such a task should come to the account. Therefore, we could use time anticipation for preliminary preparation of information so that when the task came to the account, the necessary information was already ready (magnetic tapes were twisted, and the first portion of information was transferred to RAM, etc.). To do this, a task schedule was introduced, and with the help of multi-programming, it remained only to fill in the gaps with the account irregular tasks or debugging new tasks that arise as a result of system development.

After the "Lviv system" in the late 60s - early 70s, we completed work on the "Kuntsevo" system (for the Kuntsevo radio plant). It was done in such a way as to cover almost most of the tasks in the group of instrument and machine-building industries.

We managed to sign relevant orders that 600 systems that were being developed at that time in nine defense ministries (machine-building and instrument-making) should be made on the basis of the Kuntsevo system…

• • •

Victor Mikhailovich Glushkov played a huge role in shaping the ideas of creating automated management information systems. Under his leadership, the development of special technical means for controlling a number of technological processes in the metallurgical, chemical and shipbuilding industries, as well as for microelectronics, was carried out.

Analyzing the possibilities of computer technology and the problems of economics, V.M. Glushkov was the first to put forward the idea of creating automated management information systems for enterprises as the lowest element of an automated management system for the economy as a whole.

In 1962, V.M. Glushkov gave a lecture in Lviv on the possibilities of computers. The director of the Electron plant, S.O. Petrovsky, approached him and offered his plant as a testing ground for the creation of an automated management information system by the enterprise. Thus began the development of the country's first ASU, which was named "Lviv". The first stage of the system was commissioned by the State Commission in 1967. It was recognized as a model for enterprises with a mass nature of production and replication of this system began in all branches of the defense industry. More than 1000 such systems have been implemented.

In 1967, the country's first automated management information system (ASU) of an enterprise with a mass production character "Lviv" was commissioned and recommended for mass replication. Many of the principles underlying automated management information systems of other types have been worked out on this system. In 1970, V.M. Glushkov (with a team of authors) was awarded the State Prize of the Ukrainian SSR for this development.

In parallel with the Lviv system, the systems "Project", "Speed", "Turn", "Drawing" were developed, which automated individual product life cycles.

V.M. Glushkov, together with the teams of automated management information system developers, forms a new task - the creation of complex automated management information systems.

The first such system was developed for NPO Energia (Kaliningrad, Moscow region). Within its framework, the country's first flexible production system (GIS) was created, where only automatic machines and robots worked. This was a step towards the creation of automatic control plants, which the developers of the automated management information system were already beginning to dream about. Buran was made on this GIS. The system was awarded the USSR State Prize.

The creation of the NPO "Energia" Complex ASU lasted 5 years. The system was also adopted by the State Commission. This development was led by Academician of the National Academy of Sciences of Ukraine V.I. Skurikhin and A.A. Morozov.

At the end of the 60s, the automated management information system "Kuntsevo" was created under the leadership of A.A. Stogniy, which was also recognized as a model for its class of enterprises and recommended for replication.

The work on the Lviv system allowed to form a team of automated management information system developers. Among them are Skurikhin V.I., Shkurba V.V., Morozov A.A. and a number of others.

The ideology of the automated management information system becomes generally accepted in the 70s. And when a decision is made in the USSR on the construction of the Ulyanovsk Aviation Industrial Complex, at the same time, a decision is made on the creation of the Complex ASU for this enterprise. This was the first example of designing a huge industrial complex with a built-in automated management information system. The delivery of the Ulyanovsk Aviation Industrial Complex cash register was carried out simultaneously with the release of the first Ruslan aircraft by the enterprise. The chief designer of the system was A.A. Morozov.

The experience of creating a number of automated organizational management systems allowed V.M. Glushkov to formulate the task of a nationwide automated management system (OGAS).

In our country, all organizations were poorly prepared for the perception of processing economic information. The blame lay both on the economists, who practically did not count anything, and on the creators of the computer. As a result, there was such a situation that our statistical bodies and partially planned ones were equipped with calculating and analytical machines of the 1939 model, by that time completely replaced in America by computers.

Americans developed two lines before 1965: scientific machines (these are binary floating-point machines, high-bit) and economic machines (sequential binary-decimal with advanced memory, etc.). For the first time these two lines were connected in IBM machines.

I organized a team at our institute, I myself developed a program to familiarize them with the task set by A. N. Kosygin. I spent a week in the USSR CSU, where I studied his work in detail. I looked through the entire chain from the district station to the CSU of the USSR.

In 1963, I visited at least 100 facilities, enterprises and organizations of various profiles: from factories and mines to state farms. Then I continued this work, and in ten years the number of objects reached almost a thousand. Therefore, I have a very good idea, perhaps like no one else, of the national economy as a whole: from the bottom to the very top, the features of the existing management system, the difficulties that arise and what should be considered.

The understanding of what is needed from technology, I had quite quickly. Long before the end of the introductory work, I put forward the concept of not just individual state centers, but a network of computing centers with remote access, i.e. I put modern technical content into the concept of collective use.

We have developed the first draft of theUnified State Network of Computing Centers(EGSVC), which included about 100 centers in large industrial cities and centers of economic districts connected by broadband communication channels. These centers, distributed throughout the country, in accordance with the configuration of the system, are combined with the rest engaged in processing economic information. We estimated their number at 20 thousand at that time. These are large enterprises, ministries, as well as cluster centers serving small enterprises. The presence of a distributed data bank and the possibility of unaddressed access from anywhere in this system to any information after automatic verification of the authority of the requester was characteristic. A number of issues related to the protection of information have been developed. In addition, in this two-tier system, the main computing centers exchange information among themselves not by switching channels and switching messages, as is now customary, broken down into letters, but I suggested connecting these 100 or 200 centers with broadband channels bypassing the channel-forming equipment so that information could be rewritten from a magnetic tape in Vladivostok to the tape in Moscow without reducing the speed. Then all protocols are greatly simplified and the network acquires new properties. This has not been implemented anywhere in the world yet. Our project was secret until 1977 ...

From the very beginning, I formulated the task of automating the motor function of robots. I was tasked with creating an automatic hand on a trolley that would move along the control panel of any object and switch toggle switches, turn knobs, etc., at the same time, primitive vision was added to it, which would be able to perceive only the position of the arrow of the instruments or the division of the scale. But, unfortunately, I could not find a person who would like to work with mechanics, hands. And I set this task back in 1959, when no one had mentioned robots yet. If we had good workshops, we could have been the first in the world to have a mechanical arm in 1963. Unfortunately, not everything can be done.

Synthesis of all these directions - in robot manipulators with a hand, vision and artificial speech. At the same time, we started work on recognizing the meaning of phrases in Russian, i.e. in the field of semantic networks, as it is now called. This was done by A. A. Stogniy and partly by A.A. Letichevsky, they achieved good results. A.A. Stogniy prepared good programs. According to the flow of sentences at the input, this algorithm built a semantic network, i.e. determined which words correspond to which. For example, the sentence "A chair stands on the ceiling", although grammatically correct, is semantically incorrect, etc. The rudiments of a picture of the world were made, and economical coding was invented; then A.A. Stogniy switched to the recognition of discrete images, the subject of Yu.I. Zhuravlev, and I left this case, and it died down with us. It was necessary to link it with machine translation, but again there were not enough people, and I could not deal only with semantic algorithmics. And yet, when I made a report on this topic at the IFIP Congress in Munich in 1962, it was a sensation - the Americans did not have anything like this at that time. At the same time, I was elected to the Program Committee of the International Federation for Information Processing.

Research in the field of artificial intelligence occupies a significant place in the scientific heritage of V.M. Glushkov.

Under his leadership, they were conducted on a broad front. There are also works on pattern recognition (visual, speech, language, etc.), research in the field of robotics, mathematical linguistics, information systems, etc. However, the closest problem for him, which he dealt with a lot directly himself throughout his cybernetic activity, was the automation of the search for proofs of theorems. Back in 1958, while studying A.I. Shirshov's doctoral dissertation as an opponent, V.M. Glushkov made an attempt to verify the identities found by A.I. Shirshov in rings and Lie algebras using a program on the Ural computer. He closely followed the work on the creation of logical inference search algorithms in the USSR and abroad, initiated relevant research at the Institute of Cybernetics. Under his leadership, in the early 60s, experiments were conducted on the machine implementation of the Tarski algorithm and some other algorithms for finding output in solvable theories.

The problem of evidence was associated with work on analytical calculations and their implementation in computers of the MIR series. These works were based on a solid material and technical base. In 1963, a Special Design Bureau of Mathematical Machines and Systems with a small pilot production was established at the Institute of Cybernetics of the Academy of Sciences of the Ukrainian SSR. The serial production of computers developed by the Institute of Cybernetics of the Academy of Sciences of the Ukrainian SSR, which arose on the basis of the Radiopribor plant, contributed to the organization of an independent plant of Computer Control Machines (VUM).

In addition to the work on increasing the "intelligence" of the computers being created, research was conducted in other areas during the period under review. Developing the idea of a man-machine dialogue in the automation of deductive constructions, a group of employees of the Institute of Cybernetics (A.A. Letichevsky, Yu.V. Kapitonova, Z.M. Aselderov, K.P. Vershinin, V.F. Kostyrko, etc.) led by V.M. Glushkov created a language of "practical" mathematical logic and a text processing system in this language, as close as possible to the practice of researchers in the relevant sections of modern mathematics (primarily algebra), as well as the first version of the machine algorithm of evidence. The problems of further increasing the evidentiary power of the machine part of the future dialog system were solved.

Theoretical research in the field of robotics was conducted under the leadership of V.I. Rybak. A working mock-up of an "intelligent" robot capable of visually identifying simple geometric bodies, carrying out their purposeful movement with the help of a computer-controlled "hand", etc. Theoretical studies have been conducted to improve methods of automatic speech recognition and synthesis. An experimental system for recognizing merged phrases with a dictionary of up to 300 words with a low probability of error has been created (V.A. Kovalevsky, T.P. Vintsyuk, etc.).

Under the leadership of N.M. Amosov, work continued on computer simulation of reasonable behavior. From the imitation of the activity of one person, a transition has been made to the imitation of the activity of collectives. In general, interest in the use of simulation modeling for the study of social processes has increased during this period. To this end, models with a wide range of applications have been developed and software development has begun (V.M. Glushkov et al.).

In the field of biological and medical cybernetics, research on bioelectric control of human movements continued. Multichannel bioelectric control devices of the Myoton series have been developed, which have been introduced into clinical practice, primarily for the treatment of paralysis (L. S. Aleev et al.). Together with the Kiev Research Institute of Clinical Medicine named after Academician N.D. Strazhesco develops simulation models for forecasting and management (in dialogue with a doctor) in the treatment of patients with myocardial infarction. An anamnesis automation system focused on ischemic disease has been created (V.M. Glushkov, V.A. Petrukhin, etc.).

Under the leadership of A.A. Popov, automated systems for processing medical information were created (in particular, for analyzing the function of the respiratory and cardiovascular system) and implemented in medical institutions of Yalta, Odessa, Slavyansk, Kislovodsk. An automated resort management system was developed (A.A. Stogniy, A.A. Popov, etc.). Research of biological objects and regulatory systems at the cellular and system level continued (Yu. G. Antomonov, K.A. Ivanov-Muromsky, etc.). The biomedical aspects are related to those conducted under the leadership of V.V. Pavlov studies of ergatic control systems.

By the end of the 60s, a new point of view was formed on the problem of finding evidence, the essence of which boils down to the following. First of all, it is necessary to develop a practical formal language for writing mathematical sentences and their proofs. It should be close to the natural language of mathematics and in fact represent a formalization of that part of the natural language in which books on mathematics are written. The implementation of the language of mathematics is the "evidence algorithm", which allows you to check the correctness of mathematical texts written in the language, if the proofs are sufficiently detailed, or to find gaps in them. On the basis of these tools alone, an intelligent information system was built, which allows you to accumulate knowledge and use it in the process of performing mathematical research. As for the discovery of new mathematical facts and the search for proofs of complex theorems, this should be done in an interactive mode using specialized deductive tools that are created on the basis of language, evidence algorithms and information systems.