Anomalies in output when using parallel workflow

General discussion about Elmer
Post Reply
dvlierop
Posts: 15
Joined: 16 Feb 2011, 23:21

Anomalies in output when using parallel workflow

Post by dvlierop »

Hello people,

When I use ElmerGrid to partition the mesh from the electrostatics tuturial ("TwoBalls") as follows:

Code: Select all

ElmerGrid 2 2 mesh -metis 4 -partoptim -out two-balls-mesh-partitioned -removeunused
and then calculate the solution with Elmer (using BiCGStab) and view the results in Paraview, I get a few anomalies like the one depicted in the attachment (notice the white spot on the inside of the red sphere). These anomalies occur at a single node, at the intersection of the partitions. This doesn't happen in the original tutorial, when I don't use parallelization and use only a single large mesh without any partitioning.

What is causing this? Can this be avoided somehow?

I'm using Elmer 8.4 on Ubuntu 18.04, with Paraview 5.4.1.

Thanks in advance!

Diederik
Attachments
anomaly.png
(432.25 KiB) Not downloaded yet
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Anomalies in output when using parallel workflow

Post by raback »

Hi

There was a bug related the parallel solution of capacitance matrices that has now been fixed in devel. I think the capacitance matrix has not too much been used in parallel. Thanx for finding it!

-Peter
dvlierop
Posts: 15
Joined: 16 Feb 2011, 23:21

Re: Anomalies in output when using parallel workflow

Post by dvlierop »

Thanks for fixing this Peter, within 24 hours! Amazing, keep up the good work!

Diederik
Post Reply