Application.

For each algorithm we have distinguished between different ways of running the algorithm:

``Dir''

is direct application, where we multiply with $A$ and solve with $B$.

``SI''

is shift and invert, which solves systems $(A-\sigma B)x=b$ for $x$ and multiplies with $B$. This gives the ability to compute a wider choice of eigenvalues in fewer iterations.

``Prec''

means application with a preconditioner, for instance, a sparse approximate factorization. This needs less space than shift and invert, but most often it also needs a larger number of matrix-vector multiplies.