Optimization and Parallelization of Conjugate Gradient Solver: A Survey

Author(s):

Puja Pandurang Khirodkar , Pune Institute of Computer Technology

Keywords:

Convergence, Iterative methods, Linear Systems, Parallel programming, Sparse and very large systems, Sparse Storage Formats

Abstract:

Conjugate Gradient Solver is a well-known iterative technique for solving sparse symmetric positive definite(SPD) systems of linear equations. The aim of this paper is to optimize and parallelize the currently available Conjugate Gradient Solver on GPU using CUDA which stands for Compute Unified Device Architecture, is a parallel computing platform and application programming interface (API) model created by NVIDIA. Existing Conjugate Gradient Solver can be optimized with the help of some techniques available for sparse matrix storage like Compressed Sparse Vector(CSV).CSV method uses only two arrays which helps to reduce space and time required to store the large sparse matrix. For parallelization, CUDA uses number of threads to perform large computations of iterative part of the solver.

Other Details:

Manuscript Id	:	IJSTEV2I10216
Published in	:	Volume : 2, Issue : 10
Publication Date	:	01/05/2016
Page(s)	:	691-693

Download Article