Full Title NORMA - an Extremely High-Level Language: Programming without Programs Tech Area / Field

• INF-SOF: Information and Communications / Software

Brief Description of Technology
We have developed a next-generation language called NORMA, as well as compilers for this language. We have created a very user-friendly programming language and turned from universal languages to a language that enables users to formulate problem solutions in generic terms. We are of the opinion that universal languages are only useful to system programmers. It is clear that such a language will change as soon as its application area is changed. Extreme levels of a language also mean new levels of language stability. The languages of the universal type are constantly being specified in order to reach new levels of programming convenience and to consider new computer resources. The constant modernization of the FORTRAN language is a typical example of this. The idea of creating specialized languages has existed for a long time, and now is the time to pay increased attention to this approach. We believe that this subject will soon be of major importance in computer programming.

The terminology (language) used by technical experts in complicated application areas provides us with the opportunity to consider the peculiarities of the computational environment. The notion systems in the various application areas have always been based on mathematical notions, which explains the phenomenon. It is worth mentioning that there is no “memory” notion in such languages; hence the principle of single assignment.

As the user does not have to create a program himself (this is done by a translator), the programming process itself does not cause any problems (depending, of course, on the reliability of the translator). Furthermore, apart from the traditional syntactical and semantic diagnostics, the synthesizing translator can send messages concerning errors in essential notions, e.g. on the impossibility of organizing a certain computation because of insufficient initial data or on a mistake in the index displacement. The specification of the NORMA language contains a full specification of an algorithm and does not reflect any peculiarities of the computer. NORMA may be used on computers with any kind of architecture, both sequential or parallel. The synthesizing translator must be automatically adapted to the peculiarities of the architecture or must make allowances for the particular architecture.

The high efficiency of an automatically designed program is based on the capability of deep paralleling and of making provision for the necessary level of parallelism granularity. Generally speaking, the synthesizing translator has the capability to estimate the different variants of the representation of the declarative specification in the program and to find the best one according to the built-in rule, the dialogue with the system programmer or the user himself.

Experience in the writing and measuring of 24 NORMA programs of 500 to 1500 lines each shows that a few acceptable restrictions to the NORMA programs allow the synthesis of efficient parallel programs for grid method solutions for real problems in applied mathematics, e.g. for 3-D viscous incompressible fluid flow and the simulation of a supernova explosion.

The same task specifications as in the NORMA language were used in the solutions for programming patterns with common memory and for a pattern with message passing. In the first case, we used the SPP1000 system (by CONVEX, nowadays HP, memory type NUMA), and in the second one - the multiprocessor system MVS100 based on i860 (memory type NORMA).

Legal Aspects
No patents.

Special Facilities in Use and Their Specifications
None in this research.

Scientific Papers
I.B. Zadykhailo. Organizing loop processes of computations on parameter records of special type. Journal of calcul. math. and math. physics, v3, 2 (1963), p.337-357. (in Russian).

A.N. Andrianov. The synthesis of loop computations in the NORMA language. Preprint of the Keldysh Inst. of Appl. Math., Russian Academy of Science, 171 (1986), p.28. (in Russian).

A.N. Andrianov, K.N. Efimkin, I.B. Zadykhailo. A non-procedural language for mathematical physics. Programming and Computer Software, v17, 2 (1991), pp. 121-133.

A.N. Andrianov, A.B. Bugerja, K.N. Efimkin, I.B. Zadykhailo. The specification of the NORMA language. Draft Standard. Preprint of Keldysh Ins. of Appl. Math. Russian Academy of Sc., 120 (1995), pp.1-50. (in Russian, Translation into English).

I.B.Zadykhailo, K.N.Efimkin. Meaningful terms and new generation languages (the problem of stability, friendly interface and adaptation to computational environment ensurance). Information technologies and computational systems, 2 (1996), pp. 46-58. (in Russian).

A.N.Andrianov, E.A.Andrianova. The synthesis of loop process by nonprocedural specification. Programmirovanie, 4 (1996), pp. 62-72. (in Russian).

M.M. Vasiliev, K.N. Efimkin, V.N. Ivanova. On the application of the hydrodynamic potential methods for the viscous fluid flow problem, Mathematical Modeling, v.6, 10 (1994), p. 58-65.

A.N. Andrianov, S.B. Bazarov, K.N. Efimkin. Solution of two-dimensional gas dynamics problems by the Godunov method on parallel computers using NORMA language. Preprint of Keldysh Ins. of Appl. Math., Russian Academy of Sc., 9 (1997), pp.1-24. (in Russian).

Дата добавления: 2015-07-20