FASTEig is a collection of eigensolvers that can be used to compute eigenpairs or perform spectral analysis of large sparse or structured matrices arising from a number of DOE applications. The algorithms implemented in FASTEig solvers include: implicitly restarted Lanczos/Arnoldi algorithm, a robust locally optimal block preconditioned conjugate gradient (LOBPCG) algorithm, preconditioned projected conjugate gradient (PPCG) algorithm, generalized preconditioned local harmonic projection (GPLHR), structure preserving algorithms for computing eigenpairs of linear response operators (in TDDFT or Bethe-Salpeter equations) and a spectrum slicing algorithm for computing a relatively large number of eigenpairs of large sparse or structure matrix.

Usage and application: Many of the algorithms are implemented directly in application software tools such as Quantum Espresso, DGDFT, BerkeleyGW and MFDn.


Chao Yang

Lawrence Berkeley National Laboratory, One Cyclotron Road, Mail Stop 50F, Berkeley, CA 94720