QE/EPW tips and tricks
##############################################

Last updated: Fri Jun 23, 2023 - Fixed instructions from 2023-02-20 based on Ubuntu 22.04, which was installed in all Sysnet-owned workstations in Fall 2022.

0. Compilation on local cluster
-----------------------------------

* Download the latest version from `epw-code.org <https://epw-code.org/>`_ or clone from `GitLab <https://gitlab.com/QEF/q-e>`_.
* Unpack and configure QuantumEspresso using ``tar xvfz q-e/`` then move into the new folder with ``cd q-e/``
* Make sure to load all required modules. Do the following **in order**:

  + issue ``module purge`` to start from a clean state.

    * If ``module`` command is not available, refer to `this help page <https://www.oden.utexas.edu/sysdocs/linux/ubuntu22.html#help-module-command-not-found>`_ for instructions.

  + issue ``module load intel`` to be able to load compiler modules.

  + issue ``module load compiler`` to get access to ``ifort`` and ``icc``.

  + issue ``module load mpi`` to load MPI modules.

    * As of 23 Jun 2023, neither ``mpich2`` nor ``openmpi`` is available to load.

  + issue ``module load mkl`` to load libraries.

* Issue ``./configure``
* Check ``make.inc`` file if the correct compilers are selected
* Issue ``make epw``

**NOTE:** When executing QuantumEspresso, the same module files have to be loaded as when you compiled it.


1. Issues Running QE and Solutions
-----------------------------------

* Large supercell calculations
  
  + **Memory issues:** Increase number of nodes, decrease parallelization levels, or use large-memory nodes (*nvdimm* queue on Frontera)
    
  + **Diagonalization taking too much time:** If diagonalization is taking a long time, it's possible that 'diago_thr_init' (diagonalization threshold) is too small by default, because it scales as 1/n_elec. Manually setting it to higher values can avoid such issues, but one must verify the accuracy of the resulting calculation.