Parallel Matrix Transposition and Vector Multiplication Using OpenMP

• Main Authors: Batjargal, Delgerdalai, 白德格
• Other Authors: Weng, Tienhsiung
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/50550637149183575586

碩士 === 靜宜大學 === 資訊工程學系 === 101 === In this thesis, we propose two parallel algorithms for sparse matrix-transpose and vector multiplication using CSR (Compressed Sparse Row) format. Even though this storage format is simple and hence easy to understand and maintained, one of its limitation is difficult to parallelized, and a performance of a naïve parallel algorithm can be worst. But by preprocessing useful information that is hidden and indirect in its data structure during reading a matrix from a file, our algorithm of the matrix transposition can then be performed in parallel using OpenMP. Our codes are run on a quad-core Intel Xeon64 CPU E5507 platform. We measure, and compare the performance of our algorithms with that of using Compressed Sparse Block (CSB) format. Our experimental results show that our algorithms are comparable to the CSB based algorithm when the nonzero are scatter around the matrix and size of matrix is growing.