Publication of the journal

The section is currently being updated

The parallel architectures of computing systems, including the massively parallel ones, attract a particular interest of modern researchers in the theoretical informatics field. In this connection the hardware or algorithmic acceleration of the interprocessor exchange becomes the main objective. One of the approaches to creation of algorithms can be the use of nonconventional formalism-neural networks or cellular automata (CA), to realize the model of near interaction of elementary calculators. In the work three operations with matrix data have been considered: unary, reflection, transposing. The operations have been realized by parallel algorithms in the formalism of the cellular automata in an assumption that the data had been loaded into CA before the calculation. It has been shown that all presented algorithms have linear complexity of the matrix size. Movement and modification of the data have been executed by means of introducing the bit or/and trit flag components into a cell state description. The calculation stop-conditions are the occurrence of a stop-condition of a special CA cell or of all cells of the CA field, i.e. their «freezing». The development of the cellular automata algorithmization on an example of elementary operations over matrices can be used as a base for solving more difficult tasks (for example, the calculations of a determinant of a matrix).
Mariya A. Zapletina
Institute for Design Problems in Microelectronics of Russian Academy of Sciences, Moscow, Russia

124498, Moscow, Zelenograd, Bld. 1, Shokin Square, MIET, editorial office of the Journal "Proceedings of Universities. Electronics", room 7231

+7 (499) 734-62-05
magazine@miee.ru