1: \begin{abstract}

2: Computing quark propagators in lattice QCD is equivalent to solving 

3: large, sparse linear systems with multiple right-hand sides. 

4: Block algorithms attempt to accelerate the convergence of iterative

5: Krylov-subspace methods by solving the multiple systems simultaneously.

6: This paper compares a block generalisation of the quasi-minimal residual

7: method (QMR), Block Conjugate Gradient on the normal equation, 

8: Block Lanczos and ($\gamma_5$-symmetric) Block BiConjugate Gradient.

9: %%It is found that the best block algorithms clearly outperform BiCGSTAB

10: %%on small lattices, but that the advantages diminish as the lattice volume

11: %%is increased. 

12: \end{abstract}

13: