On Mar 1, 2012, at 1:58 PM, John Fettig wrote:
> On Thu, Mar 1, 2012 at 1:32 PM, Sean Farley <sean at mcs.anl.gov> wrote:
> On Thu, Mar 1, 2012 at 12:09 PM, John Fettig <john.fettig at gmail.com> wrote:
> Sorry to dig up an old thread (I seem to have a tendency of doing that
> here!), but this problem exists in petsc-3.2-p6 with MUMPS and Metis as well.
> Looking at the differences between the dev version of mump.c and the release
> version, I don't see anything that would have obviously fixed this. Could
> you give me a hint about what was changed? Is it as simple as using the
> updated Metis package (in dev)? How difficult would it be to backport the
> fix to the released version?
>
> It'd be pretty murky to do this. Start by looking at this changeset:
>
> http://petsc.cs.iit.edu/petsc/petsc-dev/rev/0c0bf07c7a32
>
> Do I understand correctly that this was fixed mainly by using the latest
> Metis? Were there any patches to MUMPS that weren't directly related to the
> updated Metis? I'm just trying to understand what fixed the ccgraph.c
> problem. It doesn't seem to happen in 3.1-p8, although that version of MUMPS
> is a lot slower (not sure why).
>
> Related question, would it be a problem to just use petsc-dev?
>
> I could, but it's a little too unstable for use in production software.
> Using the released/patched version simplifies a lot of things on many levels.
John,
If you are using a different version of MUMPS or Metis than the petsc
version was designed for then you are still using untested, unsupported code so
you are not gaining anything. In fact you are in worse shape because we will
not support that configuration, while we will help people with problems with
petsc-dev and the latest MUMPS/Metis we support.
Barry
>
> John
>
