Tools for the Development of Memory-Efficient Sparse Linear Solvers

点击翻译按钮获取中文摘要

Tools for the Development of Memory-Efficient Sparse Linear SolversRecent advances in both hardware and algorithm development have increasedthe effect of memory access on the runtime of large scale linear algebraapplications. The PI proposes to create tools that will provide the codedeveloper with information about memory costs of codes in preparationas well as with software building blocks designed to lessen those costs.First, a framework for memory-efficient implementation of the sequences ofsparse Basic Linear Algebra Subprograms that typically constitute a sparselinear solver will be provided. Second, a collection of metrics for theprediction of memory usage of linear algebra software will be developed.Working from those metrics, a tool to automate the prediction processwill be built. Finally, keeping in mind that the performance of a codeultimately depends on how it compiles, a comprehensive study of compilersand the performance of the executables they produce will be carried out.