We present a general framework for a number of techniques based on pro
jection methods on 'augmented Krylov subspaces'. These methods include
the deflated GMRES algorithm, an inner-outer FGMRES iteration algorit
hm, and the class of block Krylov methods. Augmented Krylov subspace m
ethods often show a significant improvement in convergence rate when c
ompared with their standard counterparts using the subspaces of the sa
me dimension. The methods can all be implemented with a variant of the
FGMRES algorithm. (C) 1997 by John Wiley & Sons, Ltd.