[Trilinos-Users] creating Tpetra maps from off-processor data

[Hoemmen, Mark]

On 11/4/14, 12:00 PM, "trilinos-users-request at software.sandia.gov"
<trilinos-users-request at software.sandia.gov> wrote:

>
>
>Message: 1
>Date: Tue, 4 Nov 2014 00:47:11 +0000
>From: "Templeton, Jeremy Alan" <jatempl at sandia.gov>
>To: "trilinos-users at software.sandia.gov"
>	<trilinos-users at software.sandia.gov>
>Subject: [Trilinos-Users] creating Tpetra maps from off-processor data
>Message-ID: <D57C2643-EA3A-40F1-9BE2-AC28DCCA1D9C at sandia.gov>
>Content-Type: text/plain; charset="windows-1252"
>
>Hi all,
>
>We are trying to create importers for a particle scheme which which
>communicate on-processor data to neighboring processors.  We have seen
>some codes creating custom Zoltan_Comm_Obj?s to do this task such that
>the processor figures out which particles it needs to send to which other
>processors and executes that through the Zoltan_Comm_Obj.  We think we
>can accomplish this same task using the Teuchos comm with
>non-interrupting sends and receives.
>
>However, this seems to be a lot of effort to do something which seems
>like a common and high-level task: create a map from distributed
>information to be used with an importer.  So the question we have is: is
>there functionality within Tpetra/Teuchos which can more directly perform
>this task?

Hi Jeremy -- could you please clarify what you mean by "create a Map from
distributed information"?  Do you mean that each MPI process knows what
global indices it has, but doesn't necessarily know the other processes'
global indices?  If so, then you could just use the third (noncontiguous)
Map constructor.

mfh