Abstract
This paper presents a new hybrid solver based on the Schur complement method, in which computations are distributed between multiple CPUs and GPUs. In this solver, the Schur complement is formed either on CPUs (for small problems) or on GPUs (for large problems). The interface system is solved by a new multi-GPU algorithm implementing the conjugate gradient method with explicit preconditioning. Numerical simulations performed on a hybrid multi-core multi-GPU cluster demonstrate scalability and efficiency of the proposed algorithms.
Similar content being viewed by others
References
Agullo E, Giraud L et al (2011) Parallel hierarchical hybrid linear solvers for emerging computing platforms. Comptes Rendus Mécanique 339(2–3):96–103
Ezzatti P, Quintana-Ortí E, Remón A (2011) Using graphics processors to accelerate the computation of the matrix inverse. J Supercomput 58(3):429–437
He X, Holm M, Neytcheva M (2013) Parallel implementation of the Sherman–Morrison matrix inverse algorithm. In: Applied parallel and scientific computing, LNCS, vol 7782. Springer, pp 206–219
Kopysov S, Kuzmin I, Nedozhogin N, Novikov A, Sagdeeva Y (2013) Hybrid multi-gpu solver based on Schur complement method. In: Parallel computing technologies. Lecture Notes in Computer Science, vol 7979. Springer, Berlin Heidelberg, pp 65–79
Kopyssov S, Krasnopyorov I, Novikov A, Rytchkov V (2005) Parallel distributed object-oriented framework for domain decomposition. In: Domain decomposition methods in science and engineering, LNCSE, vol 40. Springer, pp 605–614
Toselli A, Widlund OB (2005) Domain decomposition methods—algorithms and theory. In: Springer series in computational mathematics, vol 34. Springer
Author information
Authors and Affiliations
Corresponding author
Additional information
This research is supported by RFBR (projects: 13-01-00101-a, 14-01-31066-mol_a, 14-01-00055-a), the Presidium and the Ural Branch of RAS (program N18, project 12-P-1-1005).
Rights and permissions
About this article
Cite this article
Kopysov, S., Kuzmin, I., Nedozhogin, N. et al. Scalable hybrid implementation of the Schur complement method for multi-GPU systems. J Supercomput 69, 81–88 (2014). https://doi.org/10.1007/s11227-014-1209-7
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11227-014-1209-7