MPI-CUDA sparse matrix–vector multiplication for the conjugate gradient method with an approximate inverse preconditioner | Publicación