DocumentCode :
2205281
Title :
Toward an Automatic Load Balanced Distribution Model in Conjugate Gradient Method for One-Dimensional One-Phase Oil Reservoir Simulation
Author :
Ismail, Leila ; Abou-Kassem, J.H.
Author_Institution :
Comput. Sci. & Software Eng., Pet. & Chem. Eng., United Arab Emirates Univ., Al-Ain, United Arab Emirates
fYear :
2010
fDate :
June 29 2010-July 1 2010
Firstpage :
2958
Lastpage :
2963
Abstract :
Oil reservoir simulators depend on numerical solvers to solve reservoir algebraic equations which represent the reservoir characteristics. Conjugate Gradient (CG) method is one of the most popular iterative linear solvers used. Parallelizing CG is then an essential step to increase the performance of the simulator using it. In this paper, we introduce an automatic parallel computational loading balancing model for CG. In particular, we introduce an automatic parallel load balanced distributed computational model for the matrix-vector multiplication which heavily contributes to the reduction of the complexity of the CG. Our algorithm allows an equal distribution of the CG computation on the number of workers involved in the parallel computing.
Keywords :
computational complexity; conjugate gradient methods; hydrocarbon reservoirs; iterative methods; linear algebra; matrix multiplication; parallel processing; petroleum industry; resource allocation; CG complexity reduction; automatic load balanced distribution model; automatic parallel computational loading balancing model; conjugate gradient method; iterative linear solvers; matrix vector multiplication; one dimensional one phase oil reservoir simulation; reservoir algebraic equation; Complexity theory; Computational modeling; Equations; Indexes; Load modeling; Mathematical model; Reservoirs; Conjugate Gradient (CG) Method; Distributed Systems; Oil Reservoir Simulator; Parallel Computing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer and Information Technology (CIT), 2010 IEEE 10th International Conference on
Conference_Location :
Bradford
Print_ISBN :
978-1-4244-7547-6
Type :
conf
DOI :
10.1109/CIT.2010.494
Filename :
5578502
Link To Document :
بازگشت