Need a reliable benchmark for your sparse matrix algorithm? This paper presents the Harwell-Boeing sparse matrix collection, a valuable resource for researchers working on sparse matrix problems. The collection offers a standardized set of test matrices from various scientific and engineering disciplines, enabling consistent and comparable algorithm evaluation. The test set includes problems in linear systems, least squares, and eigenvalue calculations, ranging from small counter-examples to large-scale test cases. This allows researchers to assess algorithm performance across a broad range of matrix sizes and problem types. The authors encourage other researchers to contribute test problems to the collection, promoting its continued growth and relevance as a benchmark for algorithm development. By offering a readily available set of test cases, this collection facilitates progress in sparse matrix research and applications.
This paper, published in ACM Transactions on Mathematical Software, directly aligns with the journal's focus on software tools and algorithms for mathematical problems. The presentation of a sparse matrix collection contributes directly to the development and testing of numerical software. References to other software packages and numerical methods further contextualize the paper within the field of mathematical software.