2021 04

The SLATE project added ROCm support for AMD GPU platforms to the latest releases of the SLATE and BLAS++ linear algebra libraries. BLAS++ now provides portable interfaces across both cuBLAS and rocBLAS backends for GPU BLAS, including Batched BLAS, as well as interfaces to traditional CPU BLAS. We are also developing support in BLAS++ for oneMKL for Intel GPU platforms. SLATE now uses BLAS++ as a portability layer to support both CUDA and ROCm. The repository is at https://bitbucket.org/icl/slate, and includes links to Users' and Developers' Guides and a Tutorial.