[paulzhang]

Am testing the performance of SU2 by solving the DPW6 case. In the FAMILY I UNSTRUCTURED VERSIONS OF GRIDS (CGNS Files) @ http://turbmodels.larc.nasa.gov/naca...ics_grids.html, I am able to convert all of the grids of 3-7 except for the 1 and 2 grids.

The conversion process can read CGNS successfully, but it exits after outputting the total number of hexahedra,
………
………
Loading CGNS data into SU2 data structures. Three dimensional
problem. 3670016 interior elements. 7347712 grid points. 6 surface
markers. 3670016 boundary elements in index 0 (Marker =
QUAD_bdy1). 3670016 boundary elements in index 1 (Marker =
QUAD_bdy2). 1024 boundary elements in index 2 (Marker =
QUAD_bdy3). 1024 boundary elements in index 3 (Marker =
QUAD_bdy4). 2048 boundary elements in index 4 (Marker =
QUAD_bdy5). 3584 boundary elements in index 5 (Marker =
QUAD_bdy6). 3670016 hexahedra.
………
………


My guess is these two grids are too big for one processor to handle, so I tried multi-processors, but it failed as well (got Segmentation fault). The error info is

………
………
Calling ParMETIS... Finished partitioning using ParMETIS (44562
edge cuts). Communicating partition data and creating halo layers.
3692305 interior elements including halo cells. 3692305 hexahedra.
7436964 vertices including ghost points.

----------------------- Preprocessing computations ----------------------
Setting local point and element connectivity.

----------------------- Preprocessing computations ----------------------
Setting local point and element connectivity.

----------------------- Preprocessing computations ----------------------
Setting local point and element connectivity. run: error:
cn12510: tasks 1,5,8,10,13: Segmentation fault Establishing MPI
communication patterns.

----------------------- Preprocessing computations ----------------------
………
………



Can someone tell what is the specific reason for this failure? Is there a workaround for this?

Thanks,
Paul