[MOAB-dev] boundary nodes on parallel mesh

[Grindeanu, Iulian R.]

Hi Nico,

This is a modified HelloParMOAB example,  which it dumps the "true" skin of the model

it can work in parallel, in the sense that even if you load a model in parallel
(mpiexec -np 3 HelloParMOAB <file.h5m> or
mpiexec -np 1 HelloParMOAB <file.h5m>)

 will get the same skin in skin.h5m

this example assumes for simplicity that your model has 3d elements

it first computes the skin on each local part, then filters out the faces that are shared with other parts

it puts what is left in a set, then writes in parallel the set (so each processor will write its part of the true skin)

If you want also the edges on the true skin, there is a little more work, but the idea is almost the same.

Iulian

________________________________
From: moab-dev-bounces at mcs.anl.gov [moab-dev-bounces at mcs.anl.gov] on behalf of Nico Schlömer [nico.schloemer at gmail.com]
Sent: Sunday, November 01, 2015 10:16 PM
To: Timothy Tautges
Cc: moab-dev at mcs.anl.gov
Subject: Re: [MOAB-dev] boundary nodes on parallel mesh

The issue is not understanding what's supposed to happen, but how to make it happen. I'm sure if I code this myself, I'll make mistakes. Something I could just copy-and-paste would prevent that.

--Nico

On Mon, Nov 2, 2015 at 3:35 AM Timothy Tautges <timothy.tautges at cd-adapco.com<mailto:timothy.tautges at cd-adapco.com>> wrote:

Draw an example out on paper with a ghost layer and mark shared, skin, etc and it will be clear.

On Nov 1, 2015 7:45 PM, "Nico Schlömer" <nico.schloemer at gmail.com<mailto:nico.schloemer at gmail.com>> wrote:
Yes, that would be much appreciated! (I don't understand from the thread [1] what exactly the solution is. People talk about a fork of sorts, shared and owned filtering etc. A piece of code in the FAQs would help.)

By the way, where do I find the FAQs? :)

--Nico

[1] http://lists.mcs.anl.gov/pipermail/moab-dev/2013/005509.html

On Mon, Nov 2, 2015 at 2:40 AM Timothy Tautges <timothy.tautges at cd-adapco.com<mailto:timothy.tautges at cd-adapco.com>> wrote:

Moab folks, this should be put in the faq if it isn't there already.

On Nov 1, 2015 7:38 PM, "Timothy Tautges" <timothy.tautges at cd-adapco.com<mailto:timothy.tautges at cd-adapco.com>> wrote:

I think that's still the way.

- tim

On Nov 1, 2015 7:27 PM, "Nico Schlömer" <nico.schloemer at gmail.com<mailto:nico.schloemer at gmail.com>> wrote:
Hi everyone,

As I'm diving deeper into MOAB, more questions pop up. Right now: I need to loop over the actual physical boundary of a parallel mesh and get the node IDs and their coordinates.

I found an old mailing list thread on the topic [1], but I'm not sure it was conclusive. What's the current best practice here?

Cheers,
Nico

[1] http://lists.mcs.anl.gov/pipermail/moab-dev/2013/005509.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20151102/8a81f53b/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: HelloParMOAB.cpp
Type: text/x-c++src
Size: 3701 bytes
Desc: HelloParMOAB.cpp
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20151102/8a81f53b/attachment-0001.cpp>