[Nek5000-users] genmap for 130k elements

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

Dear Vatslav,


Genmap uses recursive spectral bisection to partition the mesh into successively smaller subsets.  In principle, each cut should have minimally-sized separators so that communication is small.  In addition, the load should be balanced, meaning that the number of elements in each of the two newly-generated subsets should differ by at most 1.


A consequence of the first of these aims is that each newly-created  subdomain should be connected.  (In other words, the cut generates at most 2 separated subgraphs.)


It's easy to show that there are some small subgraphs where all of these conditions cannot be met.     It can also happen on larger subgraphs---which is where it become problematic because the communication overhead can then start to be quite large.   So, we have genmap flag cases whenever one of the new subgraphs is not connected.


In your case, the small integers indicate that the cut in question is near the bottom of the partition tree--in effect at the leaves--so all is ok.   There will be no adverse consequences.


hth


Paul


________________________________
From: Nek5000-users <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: Wednesday, March 28, 2018 11:56:05 PM
To: nek5000-users at lists.mcs.anl.gov
Subject: [Nek5000-users] genmap for 130k elements

Hi all,

Now I am trying to build the mesh with more than 130k elements, so I find out that genmap can not process properly the mesh. I catch the errors like:

' not connected  1  4  5  0 '.

But when I reduce the number of elements to ~ 129k it works great.

Is it a bug or I must use some flag ?

Regards,
Vatslav
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20180329/3e723ca7/attachment.html>