Releases: electronic-structure/SIRIUS
Releases · electronic-structure/SIRIUS
SIRIUS/7.7.1
SIRIUS/7.7.0
Changelog:
- CI/CD is now based on UENVs
- improve core generation in LAPW; in particular, MT potential is now augmented with the tail of free atom's potential
- robust search of LAPW linearization energies E_{\nu}
- constraint total unit cell magnetization for collinear case using two Fermi level approach
- fix API to set Hubbard J parameters
- fix compilation with clang-19
- fix number of auxiliary mpi grids for spherical functions symmetrization
SIRIUS/7.6.2
- contains fixes for using libxc-7
- minor code refactoring
SIRIUS/7.6.1
This is a bugfix release.
SIRIUS/7.6.0
Summary of changes:
- update main licence to BSD 3-clause and simplify licence boilerplate in the sources
- list of authors and contributors
- generate header file with API functions for Julia bindings
- direct reading of XML UPF-v2 files using pugixml library
- cleanup of tests and unit tests
- introduce real_occupation_matrix setting to enforce Hubbard matrix to be strictly real
- add pseudopotential radial grid cutoff as a parameter instead of a hardcoded 10 a.u. value
- add a code to generate smooth initial magnetisation at a slightly higher initialisation cost
- 'atom' utility is updated to generate high-accuracy LAPW species
- switch to GSL ODE solver for radial equation
- 3-rd order radial derivatives for LAPW functions
- API to set local and non-local occupancy matrices
- API to return major, minor version and revision
- issue a warning for non-optimal k-parallelization
- remove unused input parameters from the schema and move some input parameters to a more logical location
- use strong type for more indices
- iterator over split index
- save magnetisation in JSON output
Fixes:
- sirius_add_hubbard_atom_pair() contained a bug
- init_atoms_to_grid_idx() is now called only in magnetic case
- performance fix for symmetrisation of muffin-tin functions and occupation matrices
- bug-fix in IORA o1 radial integrals
- sirius_set_nonlocal_occupation_matrix(): count spin index starting from 0
- sirius_get_wave_functions(): handle spin index properly
- linear solver for metals now properly handles k and k+q occupancies
- compute and print DFT total energy and free energy separately
SIRIUS/7.5.2
Minor release:
- API function returns band gap
- Muffin-tin radii can be set from the input file
SIRIUS/7.5.1
Minor updates:
- rebuild base image for ci/cd
- add api function to compute G+q augmentation charge
- add link to discourse.group
- fix compilation error for ~memory_pool+tests
- fix scalar-relativistic radial solver
SIRIUS/7.5.0
This release contains the following updates:
- switch to C++17
- improvements to CMake build system (CMake >= 3.23 is now required) to have a better support for CUDA and ROCm
- scoped names for predefined macros
- improved spack recipe
- memory pool is now provided by Umpire; custom memory pool is removed
- bindings to DLA-F library to solve eigenvalue problem
- more logic in source code and namespace structuring
- improvements to python module
- improvements to the CI/CD setup
- fixes to linear solver interface
- new implementation of split index with iterators
- refactoring of mdarray class
- get rid of Band class in favour of pure functions
- remove optional dependency on boost::filesystem
- apply source code formatting and enable formatting check for each PR
SIRIUS/7.4.3
Internal release to test integration with CP2K
SIRIUS/7.4.0
This pre-release contains the following changes:
- update of the API to pass Fortran arrays natively
- interface with the Exciting FP-LAPW code
- add vc-sqnm library for local geometry optimisation
- fixes in the LDA+U+V implementation for non-magnetic case
- update Python module; drop Python2 support
- refactor Wave_function class; simplify and optimise wf::inner(), wf::transform() and wf::ortho() functions
- introduce output stream; now the output can be directed to stdout, file or buffer
- allow external order of G-vectors
- other minor fixes and optimisations in various parts of the code