[MOAB-dev] PBC

Thu Apr 10 14:26:20 CDT 2014

Hi Anton,
I think that what you want can be changed in your code. Moab does not need to do anything for it.

Let's take a simple case;
You can have a 1d mesh, from node 0 to node N-1, along a line, with x varying from 0 to N-1, for each node
You can repeat that , so you will have some nodes from -N to -1, and from N to 2N+1 (periodicity "N")

In your code, you just need to assemble you matrices carefully;
you will have only N degrees of freedom

nodes at positions -N, 0, and N will be mapped (assembled) to dof 0
at dof 1, you will have nodes at position -N+1, 1, N+1, and so on  (kN+1)

Your loop over nodes maybe start at position x=-1 (for one layer of ghosts), and end at N+1 , for one layer at the other end

So , from your description, at dof 0, will correspond to both input and output flow.

Of course, you will have to generalize for 3d, the same idea.

Would that work for you?

Iulian

________________________________
From: kanaev at ibrae.ac.ru [kanaev at ibrae.ac.ru]
Sent: Wednesday, April 09, 2014 4:18 PM
To: Grindeanu, Iulian R.
Subject: RE: PBC

 For example, i've got the tube with inlet surface and outlet surface. 2d meshing of inlet and outlet surfaces is equal node to node. Having PBCs set on inlet and outlet surfaces we can model an infinite flow by getting the data from outlet to the inlet of the tube. The most obvious way of implementing that is to make inlet and outlet surfaces shared, like a surface between two neighbouring processors and make a ghost layer of cells "before" inlet surface (doubling layer of cells containing outlet surface) and a ghost layer of cells "after" outlet surface (doubling layer of cells containing inlet surface).

I've looked briefly through resolve_shared_entities function and saw that shared entities are resolved by finding nodes with the same GLOBAL_ID tag on the different processors. I've tried to do the trick, did a simple brick geometry and manually changed nodes ids on the outlet to make them equal to node ids on the inlet, called resolve shared entities while loading mesh in parallel but it didn't resolve that vertices as shared))

Can you think out some way of implementing that with what moab provides?

On Mon, 7 Apr 2014 03:21:24 +0000, "Grindeanu, Iulian R." wrote:
P {margin-top:0;margin-bottom:0;}
Hi Anton,
I was on vacation last week; I will look at your problems; I do not understand what you need yet, maybe after I read more carefully
Iulian
________________________________
From: kanaev at ibrae.ac.ru [kanaev at ibrae.ac.ru]
Sent: Wednesday, April 02, 2014 3:23 AM
To: Grindeanu, Iulian R.
Subject: PBC

Hello Iulian!

I've sent a mail about PBC to moab-dev and i saw that in moab-dev Digest, Vol 65, Issue 32

Did it go there automatically or someone answered?

I can see it like this:

Today's Topics:

1. PBC (kanaev at ibrae.ac.ru<mailto:kanaev at ibrae.ac.ru>) 2. commit/MOAB: danwu: Minor changes to WriteNC.cpp (code formatting, remove some compiler warnings). Added a new test for writing variable gw (gauss weights). Running this test shows "Writer with name NC for file testgw.nc using extension nc was unsuccessful. Using default writer WriteHDF5Parallel for file testgw.nc". Will update NC writer later to make it work on variable gw. (commits-noreply at bitbucket.org<mailto:commits-noreply at bitbucket.org>) ---------------------------------------------------------------------- Message: 1 Date: Tue, 01 Apr 2014 15:34:28 -0400 From: <kanaev at ibrae.ac.ru<mailto:kanaev at ibrae.ac.ru>> To: MOAB dev <MOAB-dev at mcs.anl.gov<mailto:MOAB-dev at mcs.anl.gov>> Subject: [MOAB-dev] PBC Message-ID: <bc4cbd0d4c5454e36c766953a8694287 at mailx.ibrae.ac.ru<mailto:bc4cbd0d4c5454e36c766953a8694287 at mailx.ibrae.ac.ru>> Content-Type: text/plain; charset="utf-8" Hello MOAB-dev, What can be the way of implementing periodical boundary conditions using MOAB? As i can see, i need to make a connectivity table for opposing entities of e.g. inlet/outlet (which generally are on different partitions) Thanks Anton -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20140401/ca7d50dc/attachment-0001.html> ------------------------------ Message: 2 Date: Tue, 01 Apr 2014 21:11:38 -0000 From: commits-noreply at bitbucket.org<mailto:commits-noreply at bitbucket.org> To: moab-dev at mcs.anl.gov<mailto:moab-dev at mcs.anl.gov> Subject: [MOAB-dev] commit/MOAB: danwu: Minor changes to WriteNC.cpp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20140410/c2e249ab/attachment.html>