The ESCAPE project: Energy-efficient Scalable Algorithms for Weather Prediction at Exascale

Müller, Andreas; Deconinck, Willem; Kühnlein, Christian; Mengaldo, Gianmarco; Lange, Michael; Wedi, Nils; Bauer, Peter; Smolarkiewicz, Piotr K.; Diamantakis, Michail; Lock, Sarah-Jane; Hamrud, Mats; Saarinen, Sami; Mozdzynski, George; Thiemert, Daniel; Glinton, Michael; Bénard, Pierre; Voitus, Fabrice; Colavolpe, Charles; Marguinaud, Philippe; Zheng, Yongjun; Van Bever, Joris; Degrauwe, Daan; Smet, Geert; Termonia, Piet; Nielsen, Kristian P.; Sass, Bent H.; Poulsen, Jacob W.; Berg, Per; Osuna, Carlos; Fuhrer, Oliver; Clement, Valentin; Baldauf, Michael; Gillard, Mike; Szmelter, Joanna; O'Brien, Enda; McKinstry, Alastair; Robinson, Oisín; Shukla, Parijat; Lysaght, Michael; Kulczewski, Michał; Ciznicki, Milosz; Piątek, Wojciech; Ciesielski, Sebastian; Błażewicz, Marek; Kurowski, Krzysztof; Procyk, Marcin; Spychala, Pawel; Bosak, Bartosz; Piotrowski, Zbigniew P.; Wyszogrodzki, Andrzej; Raffin, Erwan; Mazauric, Cyril; Guibert, David; Douriez, Louis; Vigouroux, Xavier; Gray, Alan; Messmer, Peter; Macfaden, Alexander J.; New, Nick

doi:https://doi.org/10.5194/gmd-12-4425-2019

Articles | Volume 12, issue 10

https://doi.org/10.5194/gmd-12-4425-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-12-4425-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 12, issue 10

Development and technical paper

|

22 Oct 2019

Development and technical paper |

| 22 Oct 2019

The ESCAPE project: Energy-efficient Scalable Algorithms for Weather Prediction at Exascale

Andreas Müller, Willem Deconinck, Christian Kühnlein, Gianmarco Mengaldo, Michael Lange, Nils Wedi, Peter Bauer, Piotr K. Smolarkiewicz, Michail Diamantakis, Sarah-Jane Lock, Mats Hamrud, Sami Saarinen, George Mozdzynski, Daniel Thiemert, Michael Glinton, Pierre Bénard, Fabrice Voitus, Charles Colavolpe, Philippe Marguinaud, Yongjun Zheng, Joris Van Bever, Daan Degrauwe, Geert Smet, Piet Termonia, Kristian P. Nielsen, Bent H. Sass, Jacob W. Poulsen, Per Berg, Carlos Osuna, Oliver Fuhrer, Valentin Clement, Michael Baldauf, Mike Gillard, Joanna Szmelter, Enda O'Brien, Alastair McKinstry, Oisín Robinson, Parijat Shukla, Michael Lysaght, Michał Kulczewski, Milosz Ciznicki, Wojciech Piątek, Sebastian Ciesielski, Marek Błażewicz, Krzysztof Kurowski, Marcin Procyk, Pawel Spychala, Bartosz Bosak, Zbigniew P. Piotrowski, Andrzej Wyszogrodzki, Erwan Raffin, Cyril Mazauric, David Guibert, Louis Douriez, Xavier Vigouroux, Alan Gray, Peter Messmer, Alexander J. Macfaden, and Nick New

Download

Final revised paper (published on 22 Oct 2019)
Preprint (discussion started on 21 Jan 2019)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Referee Comment', Anonymous Referee #1, 25 Feb 2019
RC2: 'Comment on "The ESCAPE project: Energy-efficient Scalable Algorithms for Weather Prediction at Exascale"', Anonymous Referee #2, 22 Mar 2019
RC3: 'Review', Anonymous Referee #3, 08 Apr 2019
AC1: 'response to all referee comments', Andreas Mueller, 04 Jun 2019

Peer-review completion

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

AR by Andreas Mueller on behalf of the Authors (04 Jun 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (17 Jun 2019) by Chiel van Heerwaarden

RR by Anonymous Referee #1 (08 Jul 2019)

RR by Anonymous Referee #2 (09 Jul 2019)

Suggestions for revision or reasons for rejection

In the abstract.
“Satisfy strict service requirements in terms of time-to-solution and energy-to-solution”. This reviewer is unaware of any service agreements for energy to solution. For any given machine, there is a time-to-solution requirement. Exploiting architectures which are more energy efficient for a given time-to-solution are explored here, but that is not the same thing. Can the authors clarify what they mean or change this sentence.

Page 3 Line 29 “highly varying kernels”, What does this mean. Also the sentence is rather long and unwieldy and could benefit from restructuring.

Page 4 Line 13 Motifs is a plural of Motif. English is an elastic language whose usage does change and evolve and it maybe that motives is becoming an accepted plural of motif (I have seen this usage elsewhere). It could be confused motives meaning reasons for doing something. This reviewer suggests using Motifs for clarity.

Page 10, last sentence. “The only true solution ...” this is an odd phrase, perhaps "One approach to avoiding ….” Is better.

Page 10/11 and figure 4. The results of Michalakes et al 2015 are used to show the amount of data that has to be communicated. Is this data from model run, or is it based on scaling from a smaller run in the paper. If it is the latter, it needs to be made explicitly clear and again, if it is the latter, is it the scaling model from the Michalakes paper?

Page 11, line 11. This sentence doesn’t appear to connect with anything else and should be removed.

Page 12 figure 5.
The red open circle and green open diamond sit atop one another. This makes them hard to see and distinguish. Can this be described in the caption to make it clear. How is the operational intensity determined for a whole time-step? What was the non-optimised performance of the Kernels only and Matmult.

Page 17 figure 9a
Why is the x-axis not “node valued”? i.e. 1 and 2 are separated by 1 unit but so is 20-24 and 24-30. Figure 9b has a correct scale.

Section 3.6 What is the conclusion to the work on the comparison between processors for run-time and energy consumption?

Page 22 Line 17
The DSL allows to perform optimisations —> the DSL allows optimisations to be performed

Page 23 lines 22 and 23
Referring to papers with “like in” is rather informal language for a scientific paper.
See, or see for example is better.

Page 23 line 32
Requires to include all costs —> requires all costs to be included.

Hide

ED: Publish subject to minor revisions (review by editor) (19 Jul 2019) by Chiel van Heerwaarden

AR by Andreas Mueller on behalf of the Authors (27 Aug 2019) Author's response Manuscript

ED: Publish as is (04 Sep 2019) by Chiel van Heerwaarden

AR by Andreas Mueller on behalf of the Authors (04 Sep 2019)

Short summary

This paper presents an overview of the ESCAPE project. Dwarfs (key patterns in terms of computation and communication) are identified in weather prediction models. They are optimised for different hardware architectures. New algorithms are developed that are specifically designed for better energy efficiency and improved portability through domain-specific languages. Different numerical techniques are compared in terms of energy efficiency and performance for a variety of computing technologies.