|
Article information
2016 , Volume 21, ¹ 3, p.103-112
Chistyakov A.E., Doronchenko Y.I., Sorokin D.A.
Implementation of modelling for perturbation propagating in a massive rock on a reconfigurable computer system
The paper considers an alternative implementation for modelling of geophysical processes by means of reconfigurable computer systems (RCS), because development of new fields in hard-to-reach regions requires fast and qualitative analysis of input data. However, the developed geophysical model requires taking into account a wide range of geological peculiarities and their parameters, and becomes rather complex. Owing to adaptation of the architecture to the structure of the problem under consideration, RCSs provide a significantly higher performance, which is especially typical for tightlycoupled problems such as many geological and geophysical ones. We propose a new implementation for modelling of perturbation spreading in a massive rock with various heterogeneities, which allows effective scaling of the solution if the available RCS hardware resource is increased. Using RCSs we can significantly reduce the processing time, keeping calculation accuracy unchanged, and hence, faster and more qualitative interpretation of geophysical data can be done. In implementation for the problem of modelling, we applied the most effective approach based on integro-interpolation method within a uniform rectangular region. The approach provides the required rate of generation of the process for propagation of waves. The required accuracy of calculations, which considerably influences on the quality of interpretation, is provided by the modified alternating triangular method (ATM). An important feature of the ATM for providing smooth solution is capability to solve the problem during the less number of iterations when the geological-geophysical computational problem is ill-conditioned. Experimental research was performed on the RCS Taygeta designed on the base of up-to-date Xilinx FPGAs. With the help of the RCS high processing rate and keeping the prescribed accuracy of calculations are possible. Most importantly, we have proved that application of such technologies for implementation to such problems is really promising.
[full text]
Keywords: reconfigurable computer system, modeling of ripple disturbances in an inhomogeneous massive rock, system of linear algebraic equations, modified alternating triangular method
Author(s):
Chistyakov Alexandr Evgenyevich
Dr.
Position: Leading research officer
Office: Scientific Research Institute of Multiprocessing Computing and Control Systems SFU
Address: 347928, Russia, Taganrog
Phone Office: (8634)315491
E-mail: Cheese_05@mail.ru
SPIN-code: 1135-8843Doronchenko Yuriy Ivanovich
PhD.
Position: Deputy Director (Economy)
Office: Scientific Research Center of Supercomputers and Neurocomputers
Address: 347900, Russia, Taganrog, 106, Italyanskiy alley
Phone Office: (8634)612-111
E-mail: doronchenko@superevm.ru
Sorokin Dmitriy Anatolyevich
PhD.
Position: Head of department
Office: Scientific Research Center of Supercomputers and Neurocomputers
Address: 347900, Russia, Taganrog, 106, Italyanskiy alley
Phone Office: (8634)612-111
E-mail: jotun@inbox.ru
References:
[1]. Kvasov, I. E., Pankratov, S. A., Petrov, I. B. Computation of dynamic processes in continuous media with a crack initiated by the near-surface disturbance using grid-characteristic method. Matematicheskoe modelirovanie. 2010; 22(11):109-122. ( In Russ.)
[2]. Vladimirov, V.S. Uravneniya matematicheskoy fiziki [Equations of mathematical physics]. Uchebnik dlya fizicheskikh i mekhaniko-matematicheskikh spets.vuzov. 4-e izdanie, ispravlennoe i dopolnennoe. Moscow: Nauka; 1981: 512. ( In Russ.)
[3]. Sukhinov, A.I., Zuev, V.N., Semenistyi, V.V. Uravneniya matematicheskoy fiziki [Equations of mathematical physics]. Taganrog: TRTU; 2009: 337. ( In Russ.)
[4]. Samarskiy, A.A. Teoriya raznostnykh skhem [Theory of difference schemes]. Moscow: Nauka; 1989: 432. ( In Russ.)
[5]. Sukhinov, A.I., Chistyakov, A.E., Shishenya, A.V. Error estimation for solution of the diffusion based on the schemes with weights. Matematicheskoe modelirovanie. 2013; 25(11):53-64. ( In Russ.)
[6]. Konovalov, A.N. Metod skorejshego spuska s adaptivnym poperemenno-treugol'nym pereobuslovlivatelem. Differencial'nye uravnenija [Method of steepest decent with adaptive alternating triangular preconditioner]. 2004; 40(7):953-963. ( In Russ.)
[7]. Sukhinov, A.I., Chistyakov, A.E. Adaptive analog-SSOR iterative method for solving grid equations with nonselfadjoint operators. Matematicheskoe modelirovanie. 2012; 24(1):3-20. ( In Russ.)
[8]. Kalyaev, A.V., Levin, I.I. Modul'no-narashchivaemye mnogoprotsessornye sistemy so strukturno-protsedurnoy organizatsiey vychisleniy [Modular-scalable multiprocessor system with structural-procedural organization of calculations]. Moscow: «Yanus-K»; 2003: 380. ( In Russ.)
[9]. Kalyaev, I.A., Levin, I.I., Semernikov, E.A., Shcmoylov, V.I. Rekonfiguriruemye mul'tikonveyernye vychislitel'nye struktury. [Reconfigurable multiconveyer computer structures]. Rostov-on-Don: Izd-vo YuNTs RAN; 2008: 320. ( In Russ.)
[10]. Doronchenko, Yu.I. Metod operatsionno-grafovogo opisania odnovremennykh vychisleniy dlya mnogoprotsessornykh sistem [Method of operationally graph description of simultaneous computations for multi-processor systems]. International scientific-technology conference «Multiprocessor computer systems – 2007». Taganrog: TTI YuFU; (1):11-17. ( In Russ.)
[11]. URL: http://www.hybridmemorycube.org/ (äàòà îáðàùåíèÿ: 10.11.2015)
Bibliography link:
Chistyakov A.E., Doronchenko Y.I., Sorokin D.A. Implementation of modelling for perturbation propagating in a massive rock on a reconfigurable computer system // Computational technologies. 2016. V. 21. ¹ 3. P. 103-112
|
