Stephan Rave
Stephan Rave
We should again consider switching to 'column vectors'. We might perform automatic conversion to Fortran order when instantiating NumpyVectorArrays and/or warn about this. See https://github.com/nias-project/nias/issues/10
@lrenelt might be able to give an example where this causes a significant slowdown.
For instance, show some typical engineering applications.
This would help in understanding the necessary concepts for coupling an external solver without having to know another programming language.
Maybe add a (configurable) maximum recursion depth and/or truncate data.
Discuss aspects that are relevant for using pyMOR in an HPC context, in particular interaction with MPI.