Abstract
Performance of modern processor architectures is investigated by running a scalable computer code representing ultraparallel implementation of the multigrid method. This method is intended for solving three-dimensional parabolic and elliptic differential equations on Cartesian grids. The code is shown to be efficient on server multicore processors, on Nvidia graphics accelerators, and on the 60-core Intel Xeon Phi coprocessors of the Intel MIC architecture.
Similar content being viewed by others
References
Zhukov, V.T., Novikova, N.D., and Feodoritova, O.B., Parallel multigrid method for elliptic difference equations. Part I: Main elements of the algorithm, Preprint of the Keldysh Institute of Applied Mathematics, Moscow, 2012, no. 30. http://keldysh.ru/papers/2012/prep2012-30.pdf.
Zhukov, V.T., Novikova, N. D., and Feodoritova, O.B., Parallel multigrid method for elliptic difference equations: Anisotropic diffusion, Preprints of the Keldysh Institute of Applied Mathematics, Moscow, 2012, no. 76. http://keldysh.ru/papers/2012/prep2012-76.pdf.
Zhukov, V.T., Novikova, N.D., and Feodoritova, O.B., Parallel multigrid method for solving elliptic equations, Mat. Model., 2014, vol. 6, no. 4, pp. 425–434.
Zhukov, V.T., Novikova, N.D., and Feodoritova, O.B., Multigrid method for anisotropic diffusion equations based on adaptation of smoothers, Mat. Model., (in press).
Keldysh Institute of Applied Mathematics, Hybrid computational cluster K-100. http://www.kiam.ru/MVS/resourses/k100.html.
Intel Corporation, Intel Xeon Phi Product Family. http://www.intel.com/content/www/us/en/proces-sors/xeon/xeon-phi-detail.html.
NVIDIA Corporation, CUDA Parallel Computing Platform. http://www.nvidia.com/object/cuda-home-new.html.
Intel Developer Zone, Using the Intel MPI library on Intel Xeon Phi coprocessor systems. https://soft-ware.intel.com/articles/using-the-intel-mpi-library-on-intel-xeon-phi-coprocessor-systems.
NASA Advanced Supercomputing Division, NAS Parallel Benchmarks. https://www.nas.nasa.gov/publications/npb.html.
Krasnov, M.M. and Feodoritova, O.B., Operator library for three-dimensional discrete schemes of mathematical physics problems using GPUs with CUDA architecture, Preprint of the Keldysh Institute of Applied Mathematics, Moscow, 2013, no. 9. http://library.keldysh.ru/preprint.asp?id=2013-9.
The OpenMP API specification for parallel programming. http://www.openmp.org.
Linev, A.V., Programming for Intel Xeon Phi. http://itprojects.narfu.ru/grid/material2014/2014-GRID5-XeonPhi-Architecture.pdf.
Intel Corporation, Intel Xeon Processor E7-4860 v2. http://ark.intel.com/ru/products/75249/Intel-Xeon-Processor-E7-4860-v2-30M-Cache-2-60-GHz.
Author information
Authors and Affiliations
Corresponding author
Additional information
Original Russian Text © V.T. Zhukov, M.M. Krasnov, N.D. Novikova, O.B. Feodoritova, 2015, published in Programmirovanie, 2015, Vol. 41, No. 1.
Rights and permissions
About this article
Cite this article
Zhukov, V.T., Krasnov, M.M., Novikova, N.D. et al. Multigrid effectiveness on modern computing architectures. Program Comput Soft 41, 14–22 (2015). https://doi.org/10.1134/S0361768815010077
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S0361768815010077