[Nek5000-users] Adjacent vertices / cells

[nek5000-users at lists.mcs.anl.gov]

Correct.

Face-based vertices are shared (and are thus stored redundantly in separate elements/processors).

Same for edge and corner based vertices.

Paul

________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov [nek5000-users-bounces at lists.mcs.anl.gov] on behalf of nek5000-users at lists.mcs.anl.gov [nek5000-users at lists.mcs.anl.gov]
Sent: Monday, May 08, 2017 8:20 AM
To: nek5000-users at lists.mcs.anl.gov
Subject: Re: [Nek5000-users] Adjacent vertices / cells


Hi Paul,


Ah I see! That's very helpful, seems I was visualizing elements incorrectly. Am I right in thinking then that adjacent elements 'share' sets of points on their faces? I'm unfamiliar with these subroutines but I'll check them out. Any code that sheds more light on the way they are used would be greatly appreciated.


Cheers,


Lee

________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov <nek5000-users-bounces at lists.mcs.anl.gov> on behalf of nek5000-users at lists.mcs.anl.gov <nek5000-users at lists.mcs.anl.gov>
Sent: 08 May 2017 14:04:39
To: nek5000-users at lists.mcs.anl.gov
Subject: Re: [Nek5000-users] Adjacent vertices / cells


Hi Lee,

Just to be clear re. your original question - the (i,j,k,e) topology is point-based, not cell-based,
so one doesn't have a concept of 6 neighbors unless you are in the interior of a spectral element
or on the face of a spectral element.  At an edge or corner, you could have more or fewer than
6 neighbors.

If you instead define a cell (which would be defined, say, at the centroids of the N^3 cells within each element), you would have 6 neighbors unless you're on the boundary.

I would normally find these neighbors by invoking calls to setupds and dssum().  In one exchange, you could exchange your neighbors element number.  In a subsequent exchange you could exchange the cell index.   If you use gs_op() directly, I believe it supports integer arguments, rather than making the unseeingly an perhaps error-prone conversion to integers, though I've never had any difficulty call dssum directly and using reals.

It's important to use the gs_op (or equivalent drum) operations as that is the only viable/scalable approach in parallel. (Everything has been tested on anywhere from 1 to 6 million ranks).

I've done this several times before...and can probably dig up some code to give you cell based data,
assuming that's what you want.

Best,

Paul




________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov [nek5000-users-bounces at lists.mcs.anl.gov] on behalf of nek5000-users at lists.mcs.anl.gov [nek5000-users at lists.mcs.anl.gov]
Sent: Monday, May 08, 2017 7:38 AM
To: nek5000-users at lists.mcs.anl.gov
Subject: Re: [Nek5000-users] Adjacent vertices / cells


Hi Paul,


Thankyou very much for your prompt response,


I'm tracking an immersed object within the fluid. I can determine whether a velocity point is within the boundary using face normals and dot-products, but I need to know whether it is connected to an outside-boundary cell.


Cheers,


Lee

________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov <nek5000-users-bounces at lists.mcs.anl.gov> on behalf of nek5000-users at lists.mcs.anl.gov <nek5000-users at lists.mcs.anl.gov>
Sent: 08 May 2017 13:26:13
To: nek5000-users at lists.mcs.anl.gov
Subject: Re: [Nek5000-users] Adjacent vertices / cells


Hi Lee,

The SEM is more function-based than cell based.

My guess is that Nek might alread support the higher-level
 operation you are trying to implement.  What is your
objective?

Best, Paul


________________________________
From: nek5000-users-bounces at lists.mcs.anl.gov [nek5000-users-bounces at lists.mcs.anl.gov] on behalf of nek5000-users at lists.mcs.anl.gov [nek5000-users at lists.mcs.anl.gov]
Sent: Monday, May 08, 2017 6:52 AM
To: nek5000-users at lists.mcs.anl.gov
Subject: [Nek5000-users] Adjacent vertices / cells


Hi all,


I'm just wondering if there's a way (given a particular velocity cell's local (i,j,k,e) co-ordinates within the domain) to access the 6 adjacent cells - even if they are located inside a different element.


If so, how does this deal with cells on the boundary, where there may exist fewer than 6 adjacent cells.


Many thanks,


Lee
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20170508/b8d74483/attachment-0001.html>