Improved coefficients for polynomial filtering in ESSEX

None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None; None, None

Improved coefficients for polynomial filtering in ESSEX

Conference Paper (2017)

Author(s)

Martin Galgon (Bergische Universität Wuppertal )

Lukas Krämer (Bergische Universität Wuppertal )

Bruno Lang (Bergische Universität Wuppertal )

Andreas Alvermann (Greifswald University)

Holger Fehske (Greifswald University)

Andreas Pieper (Greifswald University)

Georg Hager (Erlangen Regional Computing Center)

Moritz Kreutzer (Erlangen Regional Computing Center)

Faisal Shahzad (Erlangen Regional Computing Center)

Gerhard Wellein (Erlangen Regional Computing Center)

Achim Basermann (Deutsches Zentrum für Luft- und Raumfahrt (DLR))

Melven Röhrig-Zöllner (Deutsches Zentrum für Luft- und Raumfahrt (DLR))

Jonas Thies (Deutsches Zentrum für Luft- und Raumfahrt (DLR))

Affiliation

External organisation

DOI related publication

https://doi.org/10.1007/978-3-319-62426-6_5 Final published version

To reference this document use

https://resolver.tudelft.nl/uuid:932198a6-9696-4f69-a671-6a32201aa4e7

More Info

expand_more

Publication Year

2017

Language

English

Affiliation

External organisation

Pages (from-to)

63-79

Publisher

Springer

ISBN (print)

9783319624242

Event

1st InternationalWorkshop on Eigenvalue Problems: Algorithms, Software and Applications in Petascale Computing, EPASA 2015 (2015-09-14 - 2015-09-16), Tsukuba, Japan

Downloads counter

204

Abstract

The ESSEX project is an ongoing effort to provide exascale-enabled sparse eigensolvers, especially for quantum physics and related application areas. In this paper we first briefly summarize some key achievements that have been made within this project. Then we focus on a projection-based eigensolver with polynomial approximation of the projector. This eigensolver can be used for computing hundreds of interior eigenvalues of large sparse matrices. We describe techniques that allow using lower-degree polynomials than possible with standard Chebyshev expansion of the window function and kernel smoothing. With these polynomials, the degree, and thus the number of matrix–vector multiplications, typically can be reduced by roughly one half, resulting in comparable savings in runtime.