Skip to content

Latest commit

 

History

History
51 lines (35 loc) · 1.61 KB

README.md

File metadata and controls

51 lines (35 loc) · 1.61 KB

Minifftw on the Leibniz Rechenzentrum (LRZ)

Naturally, you'll want to check out the official LRZ documentation at:

https://doku.lrz.de/display/PUBLIC/High+Performance+Computing

For Minifftw it is important that you keep in mind that including the right modules on LRZ systems might work quite differently depending on which cluster you build. For example, at the time this repo has been created, the fftw-mpi module on CoolMUC-3 was named fftw/3.3.8-intel-impi, whereas it is called fftw/3.3.8-intel19-impi on CoolMUC-2.

Also, keep an eye on the LRZ's environment variables which may give important information to you about the process of compiling and linking.

Changes on LRZ in the future may break this build-script, but even then it should give you a rough idea how to get things to work ;)

Build

Important Note: You should call make clean every time before you rebuild. This repo's setup-scripts are not very good at detecting changes.

Prepare a Python environment

You will need a virtual python environment which contains some central utils like distutils, numpy and so on.

$ module load python
$ conda create -n py38 python=3.8
$ source activate py38
(py38) $ conda install numpy

CoolMUC-2

To build on this cluster, run:

module load fftw/3.3.8-intel19-impi
module load python/3.6_intel
source activate YOUR-PYTHON-ENV

Then, from the main folder of this repo, just run make lrz-mpi, or make lrz if you happen to want to build without MPI support.

You'll get an importable python shared object in the build/ folder. Import it according your own design into your python programs.