2020 10

The SLATE project released the first version of its distributed, GPU-accelerated dense linear algebra library. This release includes parallel BLAS, norms, linear system solvers (LU, Cholesky, symmetric indefinite), least squares solver (QR, LQ), Hermitian eigenvalues, generalized Hermitian eigenvalues, and the singular value decomposition (SVD). SLATE uses standards such as MPI, OpenMP, BLAS, and LAPACK. Currently it works with CUDA; ports to AMD and Intel GPUs are in progress. It has Makefile, CMake, and Spack build options. The SLATE Users' Guide, Developers' Guide, and Tutorial are available. The repository is at: https://bitbucket.org/icl/slate