Optimization for a cluster with Xeon Phi accelerators of the problem of filtration flow of liquids in three-dimensional environments with double porosity

Andrey Pervunin

Abstract


The basis of this work is a previously developed set of programs designed to simulate multiphase flows in a deformable medium having pores. In order to obtain results, a parallel software package was implemented, optimized for running on an existing cluster with installed Intel Xeon Phi accelerators. In this paper, we considered various optimization methods and techniques described in this article, which are specific for this type of accelerator, and their effect on the final run time of the program. A comparison was made of various use cases of accelerators within this cluster: symmetric mode of operation and the “Offload” mode. Numerical estimates of acceleration and efficiency values are obtained in case of using a different number of cluster nodes. This paper is entirely devoted to the issue of parallel implementation of the created algorithm and optimization of the task for a computing cluster using Intel Xeon Phi accelerators. This article has a structure consisting of two sections. By their architecture, Intel Xeon Phi multi-core coprocessors are conceptual counterparts (replacement) of graphics accelerators. When using the Intel Xeon Phi multi-core coprocessor, it is possible to obtain significant acceleration when performing calculations using the right strategy. The second section of this text contains the description of the task set earlier: the system of the equations solved in this case, the methods used for its solution and the scheme of the constructed numerical algorithm are given.


Full Text:

PDF (Russian)

References


Ju.V. Perepechko, E.I. Romenskij, G.V. Reshetova, N.L. Perepechko, V.Je. Malyshkin, K.V. Kalgin, S.E. Kireev, M.B. Ostapkevich Nelinejnaja akustika i rezhimy fil'tracii v poristyh sredah // Superkomp'juternye tehnologii v nauke, obrazovanii i promyshlennosti: Al'manah / Pod redakciej akademika V.A. Sadovnichego, akademika G.I. Savina, chl.-korr. RAN Vl.V. Voevodina. – M.: Izdatel'stvo Moskovskogo universiteta, 2013. S. 119-126.

Nelinejnyj relaksacionnyj mehanizm generacii shuma fil'tracii v poristyh sredah // Izvestija Vuzov. Radiofizika, Lebedev A. V., 2018

Ob opisanii techenij slaboszhimaemoj zhidkosti v poristyh sredah pri nelinejnom zakone fil'tracii // Zhurnal "Mehanika zhidkosti i gaza" - IPMeh RAN N. E. Leont'ev, 2013

Morozov D. N., Chetverushkin B. N., Churbanova N. G., Trapeznikova M. A. Modelirovanie zadach fil'tracii na gibridnyh vychislitel'nyh sistemah // Izvestija JuFU. Tehnicheskie nauki. 2012. # 6. S. 87-91.

Vypolnenie programm na Intel Xeon Phi. Modeli organizacii vychislenij s ispol'zovaniem Intel Xeon Phi //Rezhim dostupa hpc-education.unn.ru

Godunov S.K., Zabrodin A.V., Ivanov M.Ja., Krajko A.N., Prokopov G.P. Chislennoe reshenie mnogomernyh zadach gazovoj dinamiki. M.: Nauka. 1976.


Refbacks

  • There are currently no refbacks.


Abava  Absolutech Fruct 2020

ISSN: 2307-8162