Re: [petsc-users] Multiple solves with PCMG fail

[Mark Adams]

>
> Yeah, that should work; an alternative is to defer the estimate until
> the next solve.
>

Sounds better.  As I said one of my users has a zero RHS on the first solve.