elpa

Version:

2023a.05

Category:

math

Cluster:

Vali

Author / Distributor

https://elpa.mpcdf.mpg.de/

Description

ELPA (Eigenvalue SoLvers for Petaflop-Applications) is a high-performance library designed to solve symmetric and Hermitian eigenvalue problems efficiently on distributed memory architectures.

Key features include:

  • Highly optimized direct eigensolvers

  • Support for single and double precision

  • Support for real and complex matrices

  • MPI + OpenMP parallelization

  • Backends for ScaLAPACK and OpenMP tasks

It is widely used in quantum chemistry, materials science, and electronic structure calculations.

Documentation

ELPA is a library and does not provide a standalone executable.

Common usage includes linking ELPA in your MPI+OpenMP Fortran/C projects.

Include the appropriate ELPA module and link flags:

  For Fortran:
    use elpa

  For linking:
    mpifort -I$ELPA_INC example.f90 -L$ELPA_LIB -lelpa -o example

Official documentation:
  https://elpa.mpcdf.mpg.de/documentation/

Examples/Usage

  • Load the ELPA module:

    $ module load math/ELPA/2023.05.001-intel-2023a

  • Unload the module:

    $ module unload math/ELPA/2023.05.001-intel-2023a

  • Example usage in Fortran (excerpt):

    program test_elpa
  use elpa
  implicit none
  ! Set up matrix and call ELPA solver
end program test_elpa

  • Example linking with Intel MPI compiler:

    mpifort -I$ELPA_INC -L$ELPA_LIB -lelpa mysolver.f90 -o mysolver

Installation

Source code is obtained from ELPA