(posting on behalf of Andreas)
I realized that with some of the multi-region meshes I have created over the last year or so, the mesh boundary was not showing up in my plots anymore. I investigated this yesterday and found that the boundary identifiers are zero in the M3D-C1 timeslice HDF5 files.
It seems like the mesh adaption worked fine for this case. One can see the location of the wall in terms of triangle size, but all of the boundary identifiers are zero. I believe the identifiers are written into the 7th out of 8 columns in the 'mesh' dataset inside the timeslice HDF5 files (correct me if I am wrong). For the mesh shown below, where no boundary shows up, all values in this column are zero:
In [5]: mesh.elements.shape
Out[5]: (11322, 8)
In [7]: np.nonzero(mesh.elements[:,6])
Out[7]: (array([], dtype=int64),)
In [8]: len(np.nonzero(mesh.elements[:,6])[0])
Out[8]: 0
Do we understand why this is happening? This mesh was created after pulling the latest commit from Github yesterday and recompiling, but I have seen this also with previous code versions.
I discussed this with Nate yesterday and he pointed out that there is a possibility that under these circumstances the code might not recognize the wall and could treat the whole computational domain as plasma domain (which it should not do). I have not yet tested if this is really happening, but we do need the boundary identifiers.
(posting on behalf of Andreas)
I realized that with some of the multi-region meshes I have created over the last year or so, the mesh boundary was not showing up in my plots anymore. I investigated this yesterday and found that the boundary identifiers are zero in the M3D-C1 timeslice HDF5 files.
It seems like the mesh adaption worked fine for this case. One can see the location of the wall in terms of triangle size, but all of the boundary identifiers are zero. I believe the identifiers are written into the 7th out of 8 columns in the 'mesh' dataset inside the timeslice HDF5 files (correct me if I am wrong). For the mesh shown below, where no boundary shows up, all values in this column are zero:
In [5]: mesh.elements.shape
Out[5]: (11322, 8)
In [7]: np.nonzero(mesh.elements[:,6])
Out[7]: (array([], dtype=int64),)
In [8]: len(np.nonzero(mesh.elements[:,6])[0])
Out[8]: 0
Do we understand why this is happening? This mesh was created after pulling the latest commit from Github yesterday and recompiling, but I have seen this also with previous code versions.
I discussed this with Nate yesterday and he pointed out that there is a possibility that under these circumstances the code might not recognize the wall and could treat the whole computational domain as plasma domain (which it should not do). I have not yet tested if this is really happening, but we do need the boundary identifiers.
