Benefits from using mixed precision computations in the ELPA-AEO and ESSEX-II eigensolver projects

Alvermann, Andreas and Basermann, Achim and Bungartz, Hans-Joachim and Carbogno, Christian and Ernst, Dominik and Fehske, Holger and Futamura, Yasunori and Galgon, Martin and Hager, Georg and Huber, Sarah and Huckle, Thomas and Ida, Akihiro and Imakura, Akira and Köcher, Simone and Kreutzer, Moritz and Kus, Pavel and Lang, Bruno and Lederer, Hermann and Manin, Valeriy and Marek, Andreas and Nakajima, Kengo and Nemec, Lydia and Reuter, Karsten and Rippl, Michael and Röhrig-Zöllner, Melven and Sakurai, Tetsuya and Scheffler, Matthias and Scheurer, Christoph and Shahzad, Faisal and Simoes Brambila, Danilo and Thies, Jonas and Wellein, Gerhard (2019) Benefits from using mixed precision computations in the ELPA-AEO and ESSEX-II eigensolver projects. Japanese Journal on Industrial and Applied Mathematics, 36, pp. 699-717. Springer. doi: 10.1007/s13160-019-00360-8. ISSN 0916-7005.

PDF - Preprint version (submitted draft)
1MB

Official URL: https://link.springer.com/article/10.1007/s13160-019-00360-8

Abstract

We first briefly report on the status and recent hievements of the ELPA-AEO (Eigen-value Solvers for Petaflop Applications -- Algorithmic Extensions and Optimizations) and ESSEX II (Equipping Sparse Solvers for Exascale) projects. In both collaboratory efforts, scientists from the application areas, mathematicians, and computer scientists work together to develop and make available efficient highly parallel methods for the solution of eigenvalue problems. Then we focus on a topic addressed in both projects, the use of mixed precision computations to enhance efficiency. We give a more detailed description of our approaches for benefiting from either lower or higher precision in three selected contexts and of the results thus obtained.

Item URL in elib:

https://elib.dlr.de/137929/

Document Type:

Article

Title:

Benefits from using mixed precision computations in the ELPA-AEO and ESSEX-II eigensolver projects

Authors:

Authors	Institution or Email of Authors	Author's ORCID iD
Alvermann, Andreas	Ernst-Moritz-Arndt University of Greifswald	UNSPECIFIED
Basermann, Achim	UNSPECIFIED	https://orcid.org/0000-0003-3637-3231
Bungartz, Hans-Joachim	UNSPECIFIED	UNSPECIFIED
Carbogno, Christian	UNSPECIFIED	UNSPECIFIED
Ernst, Dominik	UNSPECIFIED	UNSPECIFIED
Fehske, Holger	Ernst-Moritz-Arndt University of Greifswald	UNSPECIFIED
Futamura, Yasunori	UNSPECIFIED	UNSPECIFIED
Galgon, Martin	UNSPECIFIED	UNSPECIFIED
Hager, Georg	Erlangen Regional Computing Center	UNSPECIFIED
Huber, Sarah	UNSPECIFIED	UNSPECIFIED
Huckle, Thomas	UNSPECIFIED	UNSPECIFIED
Ida, Akihiro	UNSPECIFIED	UNSPECIFIED
Imakura, Akira	UNSPECIFIED	UNSPECIFIED
Köcher, Simone	UNSPECIFIED	UNSPECIFIED
Kreutzer, Moritz	Erlangen Regional Computing Center	UNSPECIFIED
Kus, Pavel	UNSPECIFIED	UNSPECIFIED
Lang, Bruno	UNSPECIFIED	UNSPECIFIED
Lederer, Hermann	UNSPECIFIED	UNSPECIFIED
Manin, Valeriy	UNSPECIFIED	UNSPECIFIED
Marek, Andreas	UNSPECIFIED	UNSPECIFIED
Nakajima, Kengo	UNSPECIFIED	UNSPECIFIED
Nemec, Lydia	UNSPECIFIED	UNSPECIFIED
Reuter, Karsten	UNSPECIFIED	UNSPECIFIED
Rippl, Michael	UNSPECIFIED	UNSPECIFIED
Röhrig-Zöllner, Melven	UNSPECIFIED	https://orcid.org/0000-0001-9851-5886
Sakurai, Tetsuya	UNSPECIFIED	UNSPECIFIED
Scheffler, Matthias	UNSPECIFIED	UNSPECIFIED
Scheurer, Christoph	UNSPECIFIED	UNSPECIFIED
Shahzad, Faisal	UNSPECIFIED	UNSPECIFIED
Simoes Brambila, Danilo	UNSPECIFIED	UNSPECIFIED
Thies, Jonas	UNSPECIFIED	UNSPECIFIED
Wellein, Gerhard	Erlangen Regional Computing Center	UNSPECIFIED

Date:

27 April 2019

Journal or Publication Title:

Japanese Journal on Industrial and Applied Mathematics

Refereed publication:

Yes

Open Access:

Yes

Gold Open Access:

In SCOPUS:

Yes

In ISI Web of Science:

Yes

Volume:

DOI:

10.1007/s13160-019-00360-8

Page Range:

pp. 699-717

Publisher:

Springer

ISSN:

0916-7005

Status:

Published

Keywords:

ELPA-AEO, ESSEX, eigensolver, parallel, mixed precision

HGF - Research field:

Aeronautics, Space and Transport

HGF - Program:

Space

HGF - Program Themes:

Space System Technology

DLR - Research area:

Raumfahrt

DLR - Program:

R SY - Space System Technology

DLR - Research theme (Project):

R - Vorhaben SISTEC (old)

Location:

Köln-Porz

Institutes and Institutions:

Institut of Simulation and Software Technology > High Performance Computing
Institut of Simulation and Software Technology

Deposited By:

Thies, Jonas

Deposited On:

24 Nov 2020 13:47

Last Modified:

24 Nov 2020 13:47

Repository Staff Only: item control page