[Gmsh] GhostElements Section

Doug Shi-Dong doug.shi-dong at mail.mcgill.ca
Wed Aug 29 02:00:36 CEST 2018


Hello,


I am looking at the $GhostElements section of the MSH4 file format and there seems to be inconsistencies. According to the documentation:


<$GhostElements>
  numGhostElement(int)
  elementTag(int) partition(int) GhostCellOwner(int) GhostCellOwnerPartition(int) ...
<$EndGhostElements>


Here is my current understanding and what I would expect:

  *   'partition(int)' corresponds to the partition where the REAL cell is located
  *   'GhostCellOwner(int) corresponds to the number of partitions in which element #elementTag appears as a GHOST cell
  *   'GhostCellOwnerPartition(int)' corresponds to the list of partitions in which element #elementTag appears as a GHOST cell


However, when exporting the MSH4 format there seems to be a bug such that :

  *   'partition(int) corresponds to ONE of the partitions in which the element #elementTag appears as a GHOST cell
  *   'GhostCellOwnerPartition(int)' corresponds to the partition in which the REAL cell is located


As a result, if element #elementTag appears as a GHOST cell in multiple partitions, the current implementation will have the REAL cell partition number repeated multiple times in the 'GhostCellOwnerPartition(int)'.


Therefore, it seems that there has been a mix-up between 'partition' and 'GhostCellOwnerPartition'.


I am exporting a partitioned mesh in the MSH4 format using the latest stable binary (4.0.0, August 22 2018) and have tried the latest development build 296dd75f0b29090c28165df397cac4ff922832e5.


Thank you for time.


Best Regards,


Doug
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/gmsh/attachments/20180829/e8889daa/attachment.html>


More information about the gmsh mailing list