Ginkgo
Ginkgo is a c++14 Numerical Linear Algebra toolbox for sparse problems on GPUs. Its design is based on the concept of Linear Operator allow for a simple yet powerful interface. Ginkgo’s platform portability model splits algorithm description from architecture-specific optimized kernels implemented using the platform native frameworks SYCL/DPC++ (Intel GPUs), OpenMP (Intel CPUs).
Ginkgo focuses on Software Sustainability through its development lifecycle, systematic testing, continuous integration and benchmarking, and is a member of the E4S and xSDK projects. At the same time, the Ginkgo library benefits from recent research and provides advanced functionalities such as mixed-precision algorithms or sparse batched iterative solvers.
Application Features
Multiple Sparse matrix formats, Iterative solvers & Advanced preconditioners
The Gingko Library features -
+ Basic BLAS functionality
+ Krylov solvers (GMRES, BICGSTAB, CG, IDR), Preconditioners
+ (Multigrid, ILU, ParILU, Block Jacobi, ISAI), Matrix formats
+ (CSR, COO, ELL), Other matrix operations like SpGEMM, SpGEAM,
+ transpose
+ Sparse Batched SpMV, Iterative Solvers and Preconditioners.