Abstract
A parallel algorithm for computing points on a computation front hyperplane is described. This task arises in the computation of a quantity defined on a multidimensional rectangular domain. Three-dimensional domains are usually discussed, but the material is given in the general form when the number of measurements is at least two. When the values of a quantity at different points are internally independent (which is frequently the case), the corresponding computations are independent as well and can be performed in parallel. However, if there are internal dependences (as, for example, in the Gauss-Seidel method for systems of linear equations), then the order of scanning points of the domain is an important issue. A conventional approach in this case is to form a computation front hyperplane (a usual plane in the three-dimensional case and a line in the two-dimensional case) that moves linearly across the domain at a certain angle. At every step in the course of motion of this hyperplane, its intersection points with the domain can be treated independently and, hence, in parallel, but the steps themselves are executed sequentially. At different steps, the intersection of the hyperplane with the entire domain can have a rather complex geometry and the search for all points of the domain lying on the hyperplane at a given step is a nontrivial problem. This problem (i.e., the computation of the coordinates of points lying in the intersection of the domain with the hyperplane at a given step in the course of hyperplane motion) is addressed below. The computations over the points of the hyperplane can be executed in parallel.
Similar content being viewed by others
References
Nvidia CUDA, http://www.nvidia.com/object/cuda-home-new.html.
Intel Xeon Phi, http://www.intel.com/content/www/us/en/processors/xeon/xeon-phi-detail.html.
A. N. Andrianov, K. N. Efimkin, I. B. Zadykhailo, and N. V. Podderyugina, Preprint No. 165, IPM AN SSSR (Keldysh Inst. of Applied Mathematics, USSR Academy of Sciences, Moscow, 1985).
I. B. Zadykhailo and S. P. Pimenov, Preprint No. 139, IPM AN SSSR (Keldysh Inst. of Applied Mathematics, USSR Academy of Sciences, Moscow, 1986).
A. N. Andrianov, Preprint No. 171, IPM AN SSSR (Keldysh Inst. of Applied Mathematics, USSR Academy of Sciences, Moscow, 1986).
A. N. Andrianov, Candidate’s Dissertation in Mathematics and Physics (Moscow, 1990).
A. N. Andrianov, Doctoral Dissertation in Mathematics and Physics (Moscow, 2001).
DVM system, http://www.keldysh.ru/dvm.
M. M. Krasnov and O. B. Feodoritova, Preprint No. 9, IPM RAN (Keldysh Inst. of Applied Mathematics, Russian Academy of Sciences, Moscow, 2013), http://library.keldysh.ru/preprint.asp?id=2013-9.
Gauss-Seidel method, http://ru.wikipedia.org/wiki/Metod-Gaussa-Zeidelya.
L. Lamport, “The parallel execution of DO loops,” Commun. ACM February 17(2), 83–93 (1974).
E. M. Reingold, J. Nievergelt, and N. Deo, Combinatorial Algorithms: Theory and Practice (Prentice Hall, Englewood Cliffs, N.J., 1977; Mir, Moscow, 1980).
T. C. Hu, Integer Programming and Network Flows (Addison-Wesley, Reading, Mass., 1970; Mir, Moscow, 1974).
A. O. Latsis, Parallel Data Processing (Akademiya, Moscow, 2010) [in Russian].
Author information
Authors and Affiliations
Corresponding author
Additional information
Original Russian Text © M.M. Krasnov, 2015, published in Zhurnal Vychislitel’noi Matematiki i Matematicheskoi Fiziki, 2015, Vol. 55, No. 1, pp. 145–152.
Rights and permissions
About this article
Cite this article
Krasnov, M.M. Parallel algorithm for computing points on a computation front hyperplane. Comput. Math. and Math. Phys. 55, 140–147 (2015). https://doi.org/10.1134/S0965542515010133
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S0965542515010133