Buit couldn't DofMap::reinit() copy the send_list before distributing new dofs? That is the list we want...
(Sorry, I always reply on top using my phone.)
----- Original Message -----
From: Roy Stogner <roystgnr@...>
To: Kirk, Benjamin (JSC-EG)
Cc: John Peterson <jwpeterson@...>; libmesh-users@... <libmesh-users@...>
Sent: Tue Sep 09 11:42:54 2008
Subject: Re: [Libmesh-users] Performance of EquationSystems::reinit() with ParallelMesh
On Tue, 9 Sep 2008, Kirk, Benjamin (JSC-EG) wrote:
> At one point the send_list was used to do exactly that. In fact,
> that is the only reason it exists. I need to look back and see when
> that changed. If the send list is broken I'll factor roy's stuff
> into the DofMap to fix it.
Whoa, wait, take a closer look at my patch. There's two localize
calls it was supposed to fix. In the second call, it *uses* the
DofMap send_list, because we're synchronizing data on the new, current
But in the first call, it *can't* use the DofMap send_list, because
where data needs to come from and needs to go ends up having nothing
to do with either the old partitioning's ghost nodes or the new
partitioning's ghost nodes - it has to do with how the old partitions
and new partitions overlap, and to the best of my knowledge we've
never constructed a send_list like that before.