Single-precision arithmetic in ECHAM radiation reduces runtime and energy consumption

Cotronei, Alessandro; Slawig, Thomas

doi:https://doi.org/10.5194/gmd-13-2783-2020

Articles | Volume 13, issue 6

https://doi.org/10.5194/gmd-13-2783-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-13-2783-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 13, issue 6

Development and technical paper

|

24 Jun 2020

Development and technical paper |

| 24 Jun 2020

Single-precision arithmetic in ECHAM radiation reduces runtime and energy consumption

Alessandro Cotronei and Thomas Slawig

Download

Final revised paper (published on 24 Jun 2020)
Preprint (discussion started on 17 Jan 2020)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Interesting and very comprehensive work documentation', Anonymous Referee #1, 17 Feb 2020
- AC2: 'Answer to reviewer's comments', Thomas Slawig, 17 Apr 2020
RC2: 'Accept with minor revisions', Anonymous Referee #2, 03 Mar 2020
- AC1: 'Answer to reviewers comments', Thomas Slawig, 17 Apr 2020

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Thomas Slawig on behalf of the Authors (23 Apr 2020) Author's response Manuscript

ED: Publish as is (14 May 2020) by David Topping

AR by Thomas Slawig on behalf of the Authors (22 May 2020) Author's response Manuscript

Short summary

We converted the radiation part of the atmospheric model ECHAM to single-precision arithmetic, using a step-by-step change in all modules. A small code portion still requires higher precision. The generated code can be easily changed from double to single precision and vice versa. The quality of the output of the single-precision version is comparable to observational data and the one of the original code. The runtime was reduced by 40 %, and the energy consumption could also be decreased.