Abstract
We present a performance study of variants of shifted block Krylov subspace methods for the iterative solution of multi-shifted linear systems with multiple right-hand sides given at once. All the methods can solve the whole sequence of systems simultaneously. We analyse the effect of deflated restarting to restore the superlinear convergence rate by augmenting the Krylov subspace with approximate eigenvectors recycled over the iterations, and of block size reduction strategies to handle the situation when some of the right-hand sides converge much faster than others, or the right-hand side vectors are approximately linearly dependent.