[petsc-dev] [petsc-users] Multiple solves with PCMG fail

[Mark Adams]

On Wed, Oct 22, 2014 at 9:55 AM, Jed Brown <jed at jedbrown.org> wrote:

> Mark Adams <mfadams at lbl.gov> writes:
>
> > On Wed, Oct 22, 2014 at 2:25 AM, Jed Brown <jed at jedbrown.org> wrote:
> >
> >> Barry Smith <bsmith at mcs.anl.gov> writes:
> >>
> >> >   Jed
> >> >
> >> >     There are some additional issues when the GMRES runs for zero
> >> iterations in computing eigenvalues for Cheby.
> >>
> >> Hmm, what is right?  Suggest that the user set
> >> -ksp_chebyshev_estimate_eigenvalues_random?  Automatically fall back to
> >> that?
> >>
> >>
> > Are you implying that PETSc is using the RHS to compute the spectra for
> > cheby?
>
> That is the default, yes.
>

OK, I know resistance is futile but I'll just go on the record as saying
this is not a good idea.  GAMG uses a random vector and gives this to
Cheby.  The EPSI project that I work on has a zero RHS in their first
solve.  What if you do the setup before the solve and before you have a RHS.

Also, the random vector is high frequency and should converge faster.  The
RHS can often be low frequency.

I you have very different scales in components perhaps the RHS would be
better.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20141022/f29e9928/attachment.html>