[Archive] SpisNUM interface

Message by Mike :

Hi all,
first of all, nice GUI interface to change all the parameters!
But here is my question: What is the limit of mesh elements and nodes the system can handle? I constantly get “out of array bounds” error messages obviously connected to the volume meshing when setting the SpisNum interface (Line 1115 in SpisNuminterface.py).
Can you give me the limits or how I can circumvent that problem?

Michael Klicker

Message by julien:

Hi Mike,
Currently we are more less limited up to 1Gb of RAM for 35000 nodes. The curent absolute limite of JVM seems to be about 2Gb.
For big models, it is maybe necessary to increase the memory stack of the JVM
by the -Xmx flag inside the jython script, where the JVM is called.
However, the error that you have obtained is not related to these problems. I think that this is only due to a bad setting or deployement of DataFields issued from the pre-processing phase. Typically, you have an “out of array bounds” if a DataField is missing during the deployement.
Check this and re-contact us is needed.

Message by Mike:

thanks for the info. I did some testing (had a free workstation around…), and it seems that the real limit is somewhat lower: I used the same model for all test runs, I only increased the mesh density. The uhd definitly crashes fast, and also gives occasionally a “particle in wrong tetrahedron” error. uhd2 is still running, but it looks like its going to make it. I closely watched the memory usage, they are well beneath the limits of the system (and the 990 MB jvm stack limit set on my machine).
the data below is fyi.
uhd: 3923 nodes 23611 elements (out of bounds)
uhd2: 3251 nodes 19591 elements
hd2: 1854 nodes 10960 elements
hd: 1028 nodes 6032 elements
normal: 487 nodes 2814 elements
Thanks again for the fast info, you do a good job answering my (frequent) questions…
Michael Klicker