Effektiva Lagringsmetoder för Glesa Matriser
Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Sparse matrices are often used in numerical algorithms that solve linear equation systems. Many methods for storing sparse matrices have been proposed and implemented during the years. These methods focus primarily on minimizing the total memory consumption and the time that it takes to store a sparse matrix. This report researches the available storage methods for sparse unstructured matrices. The formats that are researched and implemented are COO, CRS and ELL. The comparisons between the formats are made based on the storage memory and time for the sparse matrices with different filling ratios. A numerical algorithm has also been implemented to study the time it takes to solve a sparse matrix with one of the available storage formats, ELL. The results show that the CRS format outperform the other formats in the storage of a sparse matrix. It is concluded that there are storage methods for sparse matrices that avoid taking up unnecessary memory space, simultaneously preserving the matrix structure and doing so within a reasonable time.
Place, publisher, year, edition, pages
IdentifiersURN: urn:nbn:se:kth:diva-166582OAI: oai:DiVA.org:kth-166582DiVA: diva2:811385