GitLab

The test sets provided in the TestSets folder were too specific to
FHI-aims, so I am removing them.

These test sets were influenced by accuracy benchmark sets in the
literature, where it makes sense for multiple codes to run calculations
for the the same geometries, as the figure of merit (the predicted
physical quantities) should not vary from code to code.

For a performance/scaling benchmark set, on the other hand, different
codes and solvers require different geometries based on their
implementation details.  For example, mixing light and heavy elements
will matter more for an all-electron code than a pseudopotential code,
and vacuum spacing in surface calculations will matter more to a
planewave code than a localized code.  It makes more sense to provide
developers with the tools necessary to create their own sets of
geometries that work with their code/solver, not shove sets of
geometires on them.

Many of the "subset" directories were, in retrospect, poorly named; for
example, the 2DMaterials directory contains slab calculations, Bulk is
an ambiguous name, and there's no rationale to separate out bulk oxides
from other bulk crystals in this set.

The contents of the README were significantly modified to improve
readability and information about the file format was added.

Changes aims control.in stubs to account for use_local_index no longer needing to be set when load_balancing is set.

Made changes to FHI-aims global settings stubs to include the new ELSI solver timings file output and re-enabled load balancing, since the analytical stress tensor now supports it.

Renamed benchmark creation tool to better describe its function.  Benchmark creation script no longer attempts to create supercells of non-periodic calculations.

Test set generation script now creates supercells appropriate for the scaling dimensionality of the material.  Added additional error checking.

Added a new Keywords metadata category to assist in automated operations on the data set.  Eliminated the Purpose metadata category, because it had no purpose (go figure) and Keywords serves this function in a limited capacity.

Created prototype for test set generation script.  It does work, however the restrictions are severe (see header).

Added two new test cases.  Renamed all geometry.in.unit files to geometry.in.  I had originally intended the former to refer to calculations intended to be scaled and the later to refer to calculations that were fixed size, but in thinking about it, this is a completely arbitrary distinction (maybe someone really does want a 2x2x2 supercell of 2000 atoms of a carbon nanotube.)

Added two test cases from CASTEP.  Modified all imported test cases to include an original folder with the original data files used.