[petsc-users] question about PC_BJACOBI

[Jed Brown]

On Thu, Sep 23, 2010 at 17:07, Matthew Knepley <knepley at gmail.com> wrote:
> It is an open secret that Krylov methods are incredibly sensitive to
> orderings, especially when combined with incomplete factorization
> preconditioners.

The ordering only affects the Krylov method in the form of rounding
for dot products and such.  But as Matt says, almost every
preconditioner is significantly order-dependent.

Jed