Abstract
A numerical approach for solving gas dynamics on Cartesian grids is considered which employs an implicit time marching scheme with the matrix-free Lower-Upper Symmetric Gauss–Seidel (LU-SGS) method for solving discrete equations. Boundary conditions are treated with an embedded-boundary method. The method has two attractive features—(1) algorithmic uniformity of calculations and (2) structured memory accesses that well fit massively parallel architectures with GPU accelerators. We propose a novel CUDA+MPI computational algorithm scalable up to hundreds of GPUs and give in-depth analysis of its implementation (interoperability issues, libraries tuning).
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Jameson A, Turkel E (1981) Implicit schemes and LU decomposition. Math Comput. doi:10.2307/2007433
Lin F, Zhenghong G, Kan H, Fang X (2014) A multi-block viscous flow solver based on GPU parallel methodology. Comput Fluids. doi:10.1016/j.compfluid.2014.02.005
Wright MJ, Candler GV (1996) A data-parallel LU relaxation method for reacting viscous flows. In: Parallel computational fluid dynamics 1995. North-Holland, Amsterdam, pp 67–74
Menshov I, Kornev M (2014) Free-boundary method for the numerical solution of gas-dynamic equations in domains with varying geometry. Math Models Comput Simul. doi:10.1134/S207004821406009X
Menshov I, Nakamura Y (2004) Hybrid explicit–implicit, unconditionally stable scheme for unsteady compressible flows. AIAA J 42(3):551–559
Van Leer B (1974) Towards the ultimate conservative difference scheme II. Monotonicity and conservation combined in a second order scheme. J Comput Phys 14(4):361–370. doi:10.1016/0021-9991(74)90019-9
Godunov SK (1959) Finite difference methods for the computation of discontinuous solutions of the equations of fluid dynamics. Math USSR Sb 47:271–306
Godunov SK, Zabrodin AV, Prokopov GP (1962) A computational scheme for two-dimensional nonstationary problems of gas dynamics and calculation of the flow from a shock wave approaching a stationary state. USSR Comput Math Math Phys 1(4):1187–1219
Pavlukhin P (2013) Parallel LU-SGS numerical method implementation for gas dynamics problems on GPU-accelerated computer systems. Vestnik Lobachevsky State Univ Nizhni Novgorod 1:213–218
Kraus J (2013) An introduction to CUDA-aware MPI. https://devblogs.nvidia.com/parallelforall/introduction-cuda-aware-mpi/. Accessed 12 March 2016
NVTX Library. http://docs.nvidia.com/cuda/profiler-users-guide. Accessed 12 March 2016
Pavlukhin P, Menshov I (2015) On implementation high-scalable CFD solvers for hybrid clusters with massively-parallel architectures. In: Parallel Computing Technologies: 13th International Conference, PaCT 2015, Petrozavodsk, Proceedings. doi:10.1007/978-3-319-21909-7_42
Acknowledgments
This research was supported by the Grant No. 14-11-00872 from the Russian Scientific Fund (theoretical aspects of the Cartesian grid free boundary method), and the Grant No. 14-01-31480 from the Russian Foundation for Basic Researches (issues of parallel implementation).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Menshov, I., Pavlukhin, P. Highly scalable implementation of an implicit matrix-free solver for gas dynamics on GPU-accelerated clusters. J Supercomput 73, 631–638 (2017). https://doi.org/10.1007/s11227-016-1800-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11227-016-1800-1