[petsc-dev] DMGetLocalToGlobalMapping(), can it be collective

[Jed Brown]

On Tue, Nov 23, 2010 at 03:32, Matthew Knepley  wrote:

> I am pro collective since I can't envision depending on the lack on
> synchronization.


This stuff is all pushed, let me know if there are problems.

Jed
-- next part --
An HTML attachment was scrubbed...
URL: 

[petsc-dev] DMGetLocalToGlobalMapping(), can it be collective

[Jed Brown]

I would like to make DMDAGetISLocalToGlobalMapping() DM-level functionality
since it is an essential ingredient in the MatGetLocalSubMatrix-based
assembly, but it's not an essential part of how every conceivable DM
functions.  I would rather allocate that memory lazily so that not every
user is required to pay for it (e.g. when they don't even have a matrix),
but creating an ISLocalToGlobalMapping is a collective operation
(somewhat strangely, because only the rarely used inversion of the map is a
collective operation).  This means that either DMGetLocalToGlobalMapping
becomes a collective operation, or everyone needs to pay for it to exist?
 Thoughts?

Jed
-- next part --
An HTML attachment was scrubbed...
URL: 

[petsc-dev] DMGetLocalToGlobalMapping(), can it be collective

[Matthew Knepley]

I am pro collective since I can't envision depending on the lack on
synchronization.

  Matt

On Mon, Nov 22, 2010 at 7:06 PM, Jed Brown  wrote:

> I would like to make DMDAGetISLocalToGlobalMapping() DM-level functionality
> since it is an essential ingredient in the MatGetLocalSubMatrix-based
> assembly, but it's not an essential part of how every conceivable DM
> functions.  I would rather allocate that memory lazily so that not every
> user is required to pay for it (e.g. when they don't even have a matrix),
> but creating an ISLocalToGlobalMapping is a collective operation
> (somewhat strangely, because only the rarely used inversion of the map is a
> collective operation).  This means that either DMGetLocalToGlobalMapping
> becomes a collective operation, or everyone needs to pay for it to exist?
>  Thoughts?
>
> Jed
>



-- 
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their experiments
lead.
-- Norbert Wiener
-- next part --
An HTML attachment was scrubbed...
URL: