Skip to content
/ lsms Public

LSMS is a code for scalable first principles calculations of materials using multiple scattering theory.

License

Notifications You must be signed in to change notification settings

mstsuite/lsms

Repository files navigation

LSMS

LSMS is a code for scalable first principles calculations of materials using multiple scattering theory.

Citing LSMS

If you publish results obtained using LSMS we ask that you cite the following publications:

  • Y. Wang, G. M. Stocks, W. A. Shelton, D. M. C. Nicholson, W. M. Temmerman, and Z. Szotek. Order-n multiple scattering approach to electronic structure calculations. Phys. Rev. Lett. 75, 2867 (1995).

and if the GPU accelerated version was used, please cite additionally:

  • M. Eisenbach, J. Larkin, J. Lutjens, S. Rennich, and J. H. Rogers. GPU acceleration of the locally selfconsistent multiple scattering code for first principles calculation of the ground state and statistical physics of materials. Computer Physics Communications 211, 2 (2017).

and for calculations using Monte-Carlo simulations:

  • M. Eisenbach, C.-G. Zhou, D. M. C. Nicholson, G. Brown, J. Larkin, and T. C. Schulthess. A Scalable Method for Ab Initio Computation of Free Energies in Nanoscale Systems. Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis, ACM, New York, 64 (2009)

If you are using LSMS on a unit cell with over 1000 atoms, also cite:

  • D. M. Rogers. Overcoming the minimum image constraint using the closest point search. J. Mol. Graphics and Modelling 68, 197-205 (2016).

Installation

CMake

The current CMake can now build both LSMS and WL-LSMS. The previous build system without CMake is depreciated an might be removed in future releases.

Prerequisites

  • CMake: >=3.18
  • make
  • Autotools: (Optional) This is needed if LibXC is desired to be compiled automatically

The following libraries needs to be preinstalled

  • HDF5
  • LAPACK/BLAS or MKL
  • MPI
  • OpenMP

The following libaries can be preinstalled, but can also be installed automatically

  • LibXC
  • Lua

Examples

Here is an example of installing LSMS with GCC-10, OpenMPI and MKL. LibXC and Lua will be installed during the build.

Clone the branch with the CMake build system

git clone https://github.com/mstsuite/lsms.git

Create a build folder in a directory of your choice.

mkdir build_lsms
cd build_lsms

Run CMake. CMake will try to find all libaries automatically.

cmake <path/to/lsms/root> \
      -DCMAKE_BUILD_TYPE=Release \
      -DCMAKE_CXX_COMPILER=mpic++ \
      -DCMAKE_C_COMPILER=gcc \
      -DCMAKE_Fortran_COMPILER=gfortran \
      -DCMAKE_CXX_FLAGS="-O3 -mtune=native" \
      -DCMAKE_Fortran_FLAGS="-O3 -mtune=native -fbacktrace -cpp -fallow-argument-mismatch" \
      -DBLA_VENDOR=Intel10_64lp

For a number of systems toolchain files are provided in the toolchain directory that provide the definitions to run CMake.

cmake -DCMAKE_TOOLCHAIN_FILE=<path/to/lsms/root>/toolchain/<system-name>.cmake \
      <path/to/lsms/root>

This is just an example for creating the build-system. CMAKE_BUILD_TYPE can be either Release or Debug. This has only an effect if no compiler flags are specified. In this case cmake will preset some flags depending on the chosen build type. It is necessary to specify a C++,C and a Fortran Compiler. The C++ compiler should be preferably be specified as the compiler wrapper, because it contains all the necessary flags and path already and it will be easier for CMake to choose the right libraries. For the GNU toolchain for example, one should prefer the compiler wrapper mpic++ over g++. CMAKE_CXX_FLAGS and CMAKE_Fortran_FLAGS are the compiler flags that will be used to build the object files. OpenMP will be search automatically by CMake and corresponding flags are added automatically. LSMS also needs LAPACK and BLAS. One can specify the desired type of LAPACK with the BLA_VENDOR definition. The corresponding documentation can be found here (CMake).

Sometimes it is necessary to explicitly define the path to math libaries. In this case the automatic search has to be turned of by setting SEARCH_LAPACK or SEARCH_BLAS to OFF. The libaries has to be specified explicitly, if the search is turned off.

cmake <path/to/lsms/root> \
      -DCMAKE_BUILD_TYPE=Release \
      -DCMAKE_CXX_COMPILER=mpic++ \
      -DCMAKE_C_COMPILER=gcc \
      -DCMAKE_Fortran_COMPILER=gfortran \
      -DCMAKE_CXX_FLAGS="-O3 -mtune=native" \
      -DCMAKE_Fortran_FLAGS="-O3 -mtune=native -fbacktrace -cpp -fallow-argument-mismatch" \
      -DSEARCH_LAPACK=OFF \
      -DSEARCH_BLAS=OFF \
      -DLAPACK_LIBARIES=liblapack.so \
      -DBLAS_LIBARIES=libblas.so

The path to the LibXC and Lua libraries can be specified explicitly. In this case the libraries will not be installed automatically.

cmake <path/to/lsms/root> \
      -DCMAKE_BUILD_TYPE=Release \
      -DCMAKE_CXX_COMPILER=mpic++ \
      -DCMAKE_C_COMPILER=gcc \
      -DCMAKE_Fortran_COMPILER=gfortran \
      -DCMAKE_CXX_FLAGS="-O3 -mtune=native" \
      -DCMAKE_Fortran_FLAGS="-O3 -mtune=native -fbacktrace -cpp -fallow-argument-mismatch" \
      -Dlibxc_LIBRARIES=<path/libxc>/lib/libxc.a
      -Dlibxc_INCLUDE_DIR=<path/libxc>/include

Build all target.

cmake --build . --parallel

The LSMS binary can then be found in the build directory under bin.

About

LSMS is a code for scalable first principles calculations of materials using multiple scattering theory.

Resources

License

Stars

Watchers

Forks

Packages

No packages published