This problem only appears for very larges meshes and ranks.
I have done a bit of debugging and there are two problems:
a) the ownSample function of the FunctionSpaceshould be used to avoid that samales are writen several times.
b) a token should be used to avoid that all ranks write into the io-stream at the same time!
This problem only appears for very larges meshes and ranks. I have done a bit of debugging and there are two problems:
a) the ownSample function of the FunctionSpaceshould be used to avoid that samales are writen several times. b) a token should be used to avoid that all ranks write into the io-stream at the same time!