|
[Sponsors] |
November 8, 2017, 11:30 |
Large Grid Partitioning Issues
|
#1 |
New Member
Geoffrey Andrews
Join Date: Oct 2017
Posts: 1
Rep Power: 0 |
Hello, everyone! I am trying to run a large case using SU2 and have hit some difficulties with the grid partitioning. My grid is for a sharp cone (3D) and has 86 million cells, and I suspect that I am running into a memory issue with ParMETIS. I get a series of "PARMETIS ERROR: Poor initial vertex distribution. Processor 384 has no vertices assigned to it!" messages and then this message at the end of partitioning:
"Finished partitioning using ParMETIS (0 edge cuts). Communicating partition data and creating halo layers. 0 interior elements including halo cells. 0 vertices including ghost points." Naturally, when the case tries to run, it then gives me a segfault. I have tried running the case with different numbers of processors (512, 864, 1024) and I have also tried leaving each node with 50% of its cores idle for more available memory in each node (maintaining the same overall number of active processors) but I still seem to get the same problem. Any suggestions? |
|
Tags |
grid, parmetis, partitioning, su2 |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Large eddy simulation on unstructured grid | vanchisen | Main CFD Forum | 4 | July 6, 2009 23:44 |
MapFields to New Grid For Extreme Grid Deformations due to Body Motion | albcem | OpenFOAM | 0 | May 5, 2009 15:17 |
2d irregular grid | Remy | Main CFD Forum | 1 | December 22, 2008 05:49 |
Grid Independent Solution | Chuck Leakeas | Main CFD Forum | 2 | May 26, 2000 12:18 |
Numerical methods for discontinuous grid interfaces? | Hansong Hang | Main CFD Forum | 12 | September 16, 1998 23:26 |