Abstract:
A library of C++ templates of iterative Krylov subspace methods with
preconditioning for the solution of large sparse systems of linear
algebraic equations on modern NVIDIA graphics processors is considered.
The methods can be used on single or multiple graphics processor units
simultaneously, i.e., on the so-called hybrid computing systems on which the
computations on central processor units are combined with the computations
on graphics processor units. The library is intended for those who need to
solve large sparse systems of linear algebraic equations.
Keywords:parallel computing; graphics processors; Krylov subspace iterative methods; control volume methods; flow in porous media.