|
This article is cited in 1 scientific paper (total in 1 paper)
Dataflow computing model as a paradigm of future mainstream of software development
D. N. Zmejev, A. V. Klimov, N. N. Levchenko, A. S. Okunev, A. L. Stempkovsky Institute for Design Problems in Microelectronics, Russian Academy of Sciences, 3 Sovetskaya Str., Moscow 124365, Russian Federation
Abstract:
The article describes the parallel dataflow computing system with dynamic forming of context and the architectural features of its implementation. The dataflow computing model allows to solve the problems that arise when creating and using supercomputers. One of these problems is the difficulty of loading the rising number of functional units of cores, while remaining in the borders of traditional programming. The article describes the advantages of the proposed computing model. The article compares the paradigm of “distribution” with the traditional paradigm of “collection”. The dataflow computing model works in the paradigm of “distribution.” The description of basic features of the architecture of the “Buran” parallel dataflow computing system and its differences from the classical flow computing systems are given. The studies give hope that the proposed computing model will become in the future the main programming paradigm for large-scale parallel computing.
Keywords:
dataflow computing system; paradigm of distribution; content-addressable memory; localization of computation.
Received: 21.08.2015
Citation:
D. N. Zmejev, A. V. Klimov, N. N. Levchenko, A. S. Okunev, A. L. Stempkovsky, “Dataflow computing model as a paradigm of future mainstream of software development”, Inform. Primen., 9:4 (2015), 29–36
Linking options:
https://www.mathnet.ru/eng/ia389 https://www.mathnet.ru/eng/ia/v9/i4/p29
|
Statistics & downloads: |
Abstract page: | 474 | Full-text PDF : | 307 | References: | 58 |
|