Implementation of Iteration Methods for Solution of Linear Equation Systems in Problems of Mathematical Physics on Reconfigurable Computer Systems

Authors

  • Ilya I. Levin Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University
  • Alexey I. Dordopulo Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University
  • Andrey V. Pelipets Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University

DOI:

https://doi.org/10.14529/cmse160401

Keywords:

reconfigurable computer systems, FPGAs, numerical methods of mathematical physics, iteration parallelization, computational pipeline

Abstract

In the paper we consider unique features of implementation of iteration methods for solution of linear equation systems in problems of mathematical physics on parallel computer systems, such as geometric decomposition of the computational domain and data parallelization in sequentially performed iterations with intensive data exchange between processors and memory. Standard methods of implementation of iteration methods of solution of linear equation systems with multiple exchanges with memory and between processors, which considerably reduce the performance, require a big number of communication channels in the computer system for implementation of complex topologies and hierarchic schemes of data storage. The solution of this problem is use of multiprocessor systems with reconfigurable architecture which allow adaptation of their architecture to the structure of iteration algorithms of mathematical physics owing to iteration parallelization. In this paper we analyze implementation of the Jacobi method for the Dirichlet problem for the Laplace equation on a reconfigurable computer system. This implementation is an example which illustrates reduction of the number of external data exchange channels, which are the most critical resource of the reconfigurable computer system.

Author Biographies

Ilya I. Levin, Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University

Dep. Dir. of Science

Alexey I. Dordopulo, Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University

Head of the Laboratory

Andrey V. Pelipets, Academician A.V. Kalyaev SRI Multiprocessor Computer System at Southern Federal University

Junior Researcher

References

Самарский, А.А. Численные методы математической физики: учебное пособие / А.А. Самарский, А.В. Гулин / 2-е изд. — М : Научный мир, 2003. — 316 с. — ISBN 5-89176-196-3.

Гергель, В.П. Теория и практика параллельных вычислений / В.П. Гергель. — М.: БИНОМ, 2007. — 424 с. — ISBN 9785947746457, 9785955600963.

Asanovic K. et al. A view of the parallel computing landscape // Communications of the ACM, 2009. — Т. 52. — №. 10. — Рр. 56-67.

Каляев, И.А. Реконфигурируемые мультиконвейерные вычислительные структуры / И.А. Каляев, И.И. Левин, Е.А. Семерников, В.И. Шмойлов. — Ростов-на-Дону: Изд-во ЮНЦ РАН, 2008. — 320 с.

Virtex-6 Family Overview. Xilinx, 2015. [Электронный ресурс]. URL: http://www.xilinx.com/support/documentation/data_sheets/ds150.pdf (дата обра-щения: 16.06.2016).

7 Series FPGAs Overview. Xilinx, 2015. [Электронный ресурс]. URL: http://www.xilinx.com/support/documentation/data_sheets/ds180_7Series_Overview.pdf (дата обращения: 16.06.2016).

UltraScale Architecture and Product Overview. Xilinx, 2016. [Электронный ресурс]. URL: http://www.xilinx.com/support/documentation/data_sheets/ds890-ultrascale-overview.pdf (дата обращения: 16.06.2016).

Тихонов, А.Н. Уравнения математической физики / А.Н. Тихонов, А.А. Самарский. — М.: Наука, 1972. — 736 с.

Левин И.И., Сорокин Д.А., Мельников А.К., Дордопуло А.И. Решение задач с существенно-переменной интенсивностью потоков данных на реконфигурируемых вычислительных системах // Вестник компьютерных и информационных техно-логий. — М.: Машиностроение, 2012. — № 2. — С. 49-56.

Левин И.И., Дордопуло А.И., Каляев И.А., Гудков В.А. Высокопроизводительные реконфигурируемые вычислительные системы на основе ПЛИС VIRTEX-7 // Программная инженерия. — М.: Издательство «Новые технологии», 2014. — № 6. — С. 3-8. ISSN 2220-3397.

Published

2016-11-18

Issue

Section

Numerical Mathematics