Christoph Junghans

Results 288 comments of Christoph Junghans

I meant `~/.local/lib/python*/site-packages`.

This PR would set the default for main prefix to `~/.local` and `lib/python*/site-packages` gets appended by python resulting in `~/.local/lib/python*/site-packages`.

> @junghans What do you think about something build on this #401 I would like to keep the cmake install for spack etc.

Yeah, `fftw-mpi` is unnecessary if I recall correctly it isn't used. To use mkl you will need to utilize mkl's fftw wrappers.

https://software.intel.com/en-us/mkl-developer-reference-c-fftw3-interface-to-intel-math-kernel-library

Yeah, I have only tested double so far as well.

Maybe we should use something like [heFFTe](http://icl.utk.edu/fft/) instead of fftw, which would give us pencil decomposition plus mkl, fftw and cuda as backend for the FFTs.

It seems this related to #437 as these files pin the numpy version as well.

> 2 - As part of my problem I need to test several sizes of the atomistic region. The example provides the CG-CG interaction tables (obtained with IBI (VOTCA), I...

> > 5 - How to apply the Hamiltonian pressure and density compensation to the code? (In order to reproduce the results of DOI:[10.1140/epjst/e2016-60151-6](https://doi.org/10.1140/epjst/e2016-60151-6), Fig. 5). > > The best...