Near-global climate simulation at 1&thinsp;km resolution: establishing a performance baseline on 4888&thinsp;GPUs with COSMO 5.0

Fuhrer, Oliver; Chadha, Tarun; Hoefler, Torsten; Kwasniewski, Grzegorz; Lapillonne, Xavier; Leutwyler, David; Lüthi, Daniel; Osuna, Carlos; Schär, Christoph; Schulthess, Thomas C.; Vogt, Hannes

doi:https://doi.org/10.5194/gmd-11-1665-2018

Articles | Volume 11, issue 4

https://doi.org/10.5194/gmd-11-1665-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-11-1665-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 4

Development and technical paper

| Highlight paper

|

02 May 2018

Development and technical paper | Highlight paper |

| 02 May 2018

Near-global climate simulation at 1 km resolution: establishing a performance baseline on 4888 GPUs with COSMO 5.0

Oliver Fuhrer, Tarun Chadha, Torsten Hoefler, Grzegorz Kwasniewski, Xavier Lapillonne, David Leutwyler, Daniel Lüthi, Carlos Osuna, Christoph Schär, Thomas C. Schulthess, and Hannes Vogt

Download

Final revised paper (published on 02 May 2018)
Preprint (discussion started on 05 Oct 2017)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review of "Near-global climate simulation at 1 km resolution" by O. Fuhrer et al.', Anonymous Referee #1, 18 Dec 2017
RC2: 'review', R.W. Ford, 22 Dec 2017
AC1: 'Final response from authors', Oliver Fuhrer, 23 Jan 2018

Peer-review completion

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

AR by Oliver Fuhrer on behalf of the Authors (23 Jan 2018) Author's response Manuscript

ED: Publish subject to minor revisions (review by editor) (07 Feb 2018) by Sophie Valcke

Dear Authors,

Thank you very much for the revised version of your (very nice) paper.

However, please note that in your response you are also supposed to provide (see https://www.geoscientific-model-development.net/for_authors/submit_your_manuscript.html): "... (3) author's changes in manuscript. Regarding author's changes, a marked-up manuscript version (track changes in Word, latexdiff in LaTeX) converted into a *.pdf including the author's response must be provided." so to help the editor's work.

As I was able to track most of your changes, I am not currently asking you to provide such a marked-up manuscript, but I would like you to point me to the sections in your new manuscript that answer the following remarks from RC1:

- The discussion on why your results also apply to global weather and climate models (RC1 comment #1)

-The sentence where you specify that you don't optimise data-movement in general, also considering internode communication (RC1 comment b)

Finally, I have the following personal minor comments:

-p.4, L22: how can you write that 1500 km -a distance- corresponds to 1.4% of the earth surface? If I calculate 1500x1500 km2 (i.e. 1500 km in each direction) I get 0.4% of the earth surface.

-p.4, L 28: remove "but"

-p.5, L17: change "MeteoFrance" for "Météo France"

-p.8 in formula (1), please specify what is "N/4"

-p. 13, caption Figure 6; I think "Strong scalability on Piz Daint: on Tesla P100 GPUs (filled symbols), and on Haswell CPUs using 12 MPI ranks per node (empty symbols)" would be clearer

-p.14, L7; remove one "on CPUS" in "...execution on CPUs yields to a shorter time to solution on CPUs than on GPUs."

With best regards,
Sophie Valcke

Hide

AR by Oliver Fuhrer on behalf of the Authors (07 Feb 2018) Author's response Manuscript

ED: Publish subject to technical corrections (08 Feb 2018) by Sophie Valcke

AR by Oliver Fuhrer on behalf of the Authors (09 Feb 2018) Author's response Manuscript

Short summary

The best hope for reducing long-standing uncertainties in climate projections is through increasing the horizontal resolution of climate models to the kilometer scale. We establish a baseline of what it would take to do such simulations using an atmospheric model that has been adapted to run on a supercomputer accelerated with graphics processing units. To our knowledge this represents the first production-ready atmospheric model being run entirely on accelerators on this scale.