Running anything in a cluster environment can cause some unexpected problems. With these notes, one should be able to run qpms successfully on Aalto's Triton.

The main problem is that it is often not completely trivial to ensure the mutual compatibility of various libraries and tools on which qpms depends.

Triton+anaconda

Installing qpms with anaconda

This section shows how to use qpms in an anaconda virtual environment. First, we purge the modules to ensure that no other other conflicting python instances are loaded. Then we load an anaconda3 module.

module purge

module load anaconda3

Next, conda needs to infect our shell with some code in order to work properly. (If you use a different shell than bash, modify the following command accordingly.)

conda init bash

This adds several lines to ~/.bashrc (or elsewhere, depending on the shell). Then we might be required to restart the shell.

Next, we create and activate a virtual environment to be used with qpms. It will be called trqpms throughout this document. We will also need to have cython installed before even trying to install qpms.

conda create trqpms
conda activate trqpms
conda install cython

Moreover, we need to load the GCC and gsl modules. GCC has to be loaded after loading anaconda3 and activating the conda virtual environment, otherwise anaconda will try to use its own (likely broken) compiler when installing qpms. And we also need cmake for building amos. Lastly, we save the list of loaded modules for later use as qpms_conda.

module load gsl
module load GCC
module load cmake
module save qpms_conda

From now on, we need to be in the qpms source root directory.

cd /path/to/qpms

Now it's time to build amos. First, we clean all the files previously created by cmake (if needed). This is to ensure that cmake uses the same gfortran that is in the current PATH; otherwise, cmake might use some Fortran compiler detected earlier (cmake apparently does not update that information) and that might later cause mismatch between libgfortran versions.

rm -r CMakeFiles CMakeCache.txt
cmake .
make clean
make amos

Cmake builds in other directories are not supported right now, as qpms setup scripts now expects certain amos files at fixed relative paths. Therefore the . in cmake ..

And finaly, build qpms.

python3 setup.py install

At this point, qpms should be installed successfully. If not, this document needs updates.

Running qpms with anaconda

After installing the qpms python module, one should be able to import it and to use the scripts in the misc directory in the same shell session used for installing qpms.

To use qpms also later, one has to restore the environment, i.e. load the same modules and activate the anaconda virtual environment.

module restore qpms_conda

conda activate trqpms