Full text of article: Russian language.

Abstract:
The work is concerned with improving a familiar algorithm for solving triangular systems on a processor ring. With the approach proposed, the low-efficient final part of the algorithm is realized sequentially. The relevance of the technique developed is proved by the results of the computing experiment.

Key words:
triangular shaped matrix, parallel algorithm, processor ring.

References:

1. Golub, G.H. Matrix Calculations / G.H. Golub, Ch.F. Van Loan – Moscow: Mir, 1999. – 548 p. – (in Russian).
2. Kosarev V.I. 12 lectures on calculus mathematics. An introduction course// V.I. Kosarev - Fizmatlit, 2000. – 224 p. – (in Russian).
3. Samarskiy A.A. Numerical methods (the Manual for high schools) / A.A. Samarskiy, A.V. Gulin - Moscow: Nauka, 1989.- 432 p. – (in Russian).
4. Methods of Computer Optics (Secondary Edition) / edited by V.A. Soifer – Moscow: Fizmatlit, 2003. – 688 с. – (in Russian).
5. Li G. A parallel triangular solver for a distributed-memory multiprocessor / G. Li and T. Coleman // J. Sci and Stat. Comp. 1998.- N9.- p.485-502.
6. Ortega D.M.  Introduction in parallel and vector methods of the decision of linear systems / D.M. Ortega – Moscow: Mir, 1991. – 364 p. – (in Russian).
7. Valkovskiy V.A. Elements of parallel programming/ V.A. Valkovskiy, V.E. Kotov, A.G. Marchuk, N.N. Mirenkov - Moscow: Radio i svyaz, 1983.- 239p. – (in Russian).
8. Iliin V.P. Problems of high-efficiency technologies of the decision of the big rarefied systems / V.P. Ilin // Computing methods and programming.- 2009.- V. 10б № 1.- p. 130-136. – (in Russian).
9.  CUBLAS Library, Published by NVIDIA Corporation 2701 San Tomas Expressway Santa Clara, CA 95050, http://www.nvidia.ru/object/cuda_get_ru.html.

---

© 2009, IPSI RAS
151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: journal@computeroptics.ru ; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20