General Linear Algebra Kernels/Solvers
Welcome to the GPU Linear Algebra Community!
This community focuses on the development, usage, and performance of GPU-accelerated linear algebra. Linear algebra has been a hot topic of research and development, as the solvers are both interesting and challenging to develop while also being useful in a broad range of applications.
My entry to this topic was in 2005, starting with electromagnetics solvers; specifically LU decomposition and GMRES for the solution of Method of Moments problems. From there, our group has expanded into general linear algebra as well as into computational fluid dynamics and image processing applications.
Here, as a community, we can discuss the latest research in the field, applications of linear algebra techniques, and methods of mapping these to the GPU for the best possible performance. We will focus on all levels, from the simplest kernels (matrix-vector products, dot products, BLAS calls) to higher complexity methods (LU, QR, eigenvalues) in both dense and sparse formats.
John Humphrey, EM Photonics