Sparse Matrix Storage new format and its efficiency

Author: Koba Gelashvili
Co-authors: Vakhtang Laluashvili, Giga Chalauri
Keywords: Sparse matrix, Conjugate gradient method, Compressed matrix - row major orientation
Annotation:

To represent a sparse matrix, we store its non-zero elemnts in dinamical array. We also store matrix of indexes,which estableshes relationship between new format and traditional, dence matrix. Unlike usual efficient formats of sparse matrix storing, We are using more memory, but not more the 2-times. Though, with respect to dense matrix, new format is still much more effective. to benchmark efficiency of new fomrat with existing effective formats, new format is compared with sparse matrix of "boost" library. The conjugate gradient method is used for quadratic functional. Corresponding program is tested on 70 sparse matrixes, taken from internet. Changing the data structure, on allmost all tests new format is significally fast. Corresponding project is done in C++. The code of conjugate gradient method is very simple. Parts of programm, responsible on proccesing collection of tests and storing data are complicated and cumbersome