[buckngnr11]

I'm having an issue I haven't found a solution to. I'm doing dynamic meshing, and have a UDF to calculate values at every wall node. Part of this code requires summing some scalar quantities stored in the faces that border each node. This works, but I noticed when I plotted contours of the N_UDMI, their are weird lines that cut through the contour with small contour values. Turns out these are the thread partions in parallel mode. If I run in serial, these disappear. I do need to run in parallel mode. What's happening is I set N_UDMI(v,0) and it's only summing the faces bordring that node that lie inside that partition of the mesh. So roughly half of the faces are summed, then the rest are summed separately on another processor and the N_UDMI is at that node is overwritten. Is there an easy way to sum these N_UDMI values that are at shared nodes across different processors?

[buckngnr11]

Alternatively, is there a way to repartition in the GUI onto one processor just to run my UDF? Then for my calculations I can repartition to multiple processors?

[vinerm]

If the UDF is not required to be executed while the case is running, then you can load case in serial mode and run UDF. Manual partitioning is also available under Parallel > Partition.

As far as values in UDM are concerned, that depends on type of loop you are using. Usually, nodes cannot overwrite values at the address that belong to other nodes, but as stated above, that depends on type of loop.