ABSTRACT
The next generation of weather and climate models will have an unprecedented level of resolution and model complexity, and running these models efficiently will require taking advantage of future supercomputers and heterogeneous hardware.
In this paper, we investigate the use of mixed-precision hardware that supports floating-point operations at double-, single- and half-precision. In particular, we investigate the potential use of the NVIDIA Tensor Core, a mixed-precision matrix-matrix multiplier mainly developed for use in deep learning, to accelerate the calculation of the Legendre transforms in the Integrated Forecasting System (IFS), one of the leading global weather forecast models. In the IFS, the Legendre transform is one of the most expensive model components and dominates the computational cost for simulations at a very high resolution.
We investigate the impact of mixed-precision arithmetic in IFS simulations of operational complexity through software emulation. Through a targeted but minimal use of double-precision arithmetic we are able to use either half-precision arithmetic or mixed half/single-precision arithmetic for almost all of the calculations in the Legendre transform without affecting forecast skill.
- Peter Bauer, Alan Thorpe, and Gilbert Brunet. 2015. The quiet revolution of numerical weather prediction. Nature 525, 7567 (2015), 47--55.Google Scholar
- Matthew Chantry, Tobias Thornes, Tim Palmer, and Peter Düben. 2019. Scale-Selective Precision for Weather and Climate Forecasting. Monthly Weather Review 147, 2 (2019), 645--655.Google ScholarCross Ref
- Andrew Dawson and Peter D. Düben. 2017. Rpe v5: An emulator for reduced floating-point precision in large numerical simulations. Geoscientific Model Development 10, 6 (2017), 2221--2230.Google ScholarCross Ref
- Peter D. Düben, Hugh McNamara, and T. N. Palmer. 2014. The use of imprecise processing to improve accuracy in weather & climate prediction. J. Comput. Phys. 271 (2014), 2--18. Google ScholarDigital Library
- Oliver Fuhrer, Tarun Chadha, Torsten Hoefler, Grzegorz Kwasniewski, Xavier Lapillonne, David Leutwyler, Daniel Lüthi, Carlos Osuna, Christoph Schär, Thomas C. Schulthess, and Hannes Vogt. 2018. Near-global climate simulation at 1km resolution: Establishing a performance baseline on 4888 GPUs with COSMO 5.0. Geoscientific Model Development 11, 4 (2018), 1665--1681.Google ScholarCross Ref
- Sam Hatfield, Peter Düben, Matthew Chantry, Keiichi Kondo, Takemasa Miyoshi, and Tim Palmer. 2018. Choosing the optimal numerical precision for data assimilation in the presence of model error. Journal of Advances in Modeling Earth Systems (2018).Google Scholar
- Sam Hatfield, Aneesh Subramanian, Tim Palmer, and Peter Düben. 2018. Improving weather forecast skill through reduced precision data assimilation. Monthly Weather Review 146 (2018), 49--62.Google ScholarCross Ref
- Nicholas J Higham. 2002. Accuracy and Stability of Numerical Algorithms. 1---663 pages. arXiv:arXiv:1011.1669v3 Google ScholarDigital Library
- Bryan N. Lawrence, Michael Rezny, Reinhard Budich, Peter Bauer, Jörg Behrens, Mick Carter, Willem Deconinck, Rupert Ford, Christopher Maynard, Steven Mullerworth, Carlos Osuna, Andrew Porter, Kim Serradell, Sophie Valcke, Nils Wedi, and Simon Wilson. 2017. Crossing the Chasm: How to develop weather and climate models for next generation computers? Geoscientific Model Development Discussions September (2017), 1--36.Google Scholar
- Martin Leutbecher. 2018. Ensemble size: How suboptimal is less than infinity? Quarterly Journal of the Royal Meteorological Society June (2018), 1--22.Google Scholar
- Paulius Micikevicius, Sharan Narang, Jonah Alben, Gregory Diamos, Erich Elsen, David Garcia, Boris Ginsburg, Michael Houston, Oleksii Kuchaiev, Ganesh Venkatesh, and Hao Wu. 2017. Mixed Precision Training. CoRR abs/1710.0 (2017). arXiv:1710.03740 http://arxiv.org/abs/1710.03740Google Scholar
- 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 Piotrowski, Andrzej Wyszogrodzki, Erwan Raffin, Cyril Mazauric, David Guibert, Louis Douriez, Xavier Vigouroux, Alan Gray, Peter Messmer, Alexander J. Macfaden, and Nick New. 2019. The ESCAPE project: Energy-efficient Scalable Algorithms for Weather Prediction at Exascale. Geoscientific Model Development Discussions January (2019), 1--50.Google Scholar
- NVIDIA. 2017. NVIDIA Tesla V100 GPU Architecture. Technical Report. http://www.nvidia.com/content/gated-pdfs/Volta-Architecture-Whitepaper-v1.1.pdfGoogle Scholar
- T. N. Palmer, R. Buizza, F. Doblas-Reyes, T. Jung, Martin Leutbecher, G. Shutts, M. Steinheimer, and Antje Weisheimer. 2009. Stochastic Parametrization and Model Uncertainty. ECMWF Tech. Memo. 598 (2009), 42. https://www2.physics.ox.ac.uk/sites/default/files/2011-08-15/techmemo598{_}stochphys{_}2009{_}pdf{_}50419.pdfGoogle Scholar
- A. J. Simmons, D. M. Burridge, M. Jarraud, C. Girard, and W. Wergen. 1989. The ECMWF medium-range prediction models development of the numerical formulations and the impact of increased resolution. Meteorology and Atmospheric Physics 40, 1-3 (1989), 28--60.Google ScholarCross Ref
- Tobias Thornes, Peter Düben, and Tim Palmer. 2017. On the use of scale-dependent precision in Earth System modelling. Quarterly Journal of the Royal Meteorological Society 143, 703 (2017), 897--908.Google ScholarCross Ref
- N. P. Wedi. 2014. Increasing horizontal resolution in numerical weather prediction and climate simulations: illusion or panacea? Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 372, 2018 (2014), 20130289--20130289.Google ScholarCross Ref
- Nils P. Wedi, Mats Hamrud, and George Mozdzynski. 2013. A Fast Spherical Harmonics Transform for Global NWP and Climate Models. Monthly Weather Review 141, 10 (2013), 3450--3461.Google ScholarCross Ref
- Hisashi Yashiro, Masaaki Terai, Ryuji Yoshida, Shin-ichi Iga, Kazuo Minami, and Hirofumi Tomita. 2016. Performance Analysis and Optimization of Nonhydrostatic ICosahedral Atmospheric Model (NICAM) on the K Computer and TSUBAME2.5. In Proceedings of the Platform for Advanced Scientific Computing Conference. ACM Press, New York, New York, USA, 1--8. Google ScholarDigital Library
Index Terms
- Accelerating High-Resolution Weather Models with Deep-Learning Hardware
Recommendations
Impact of Vectorization Over 16-bit Data-Types on GPUs
PARMA-DITAM '18: Proceedings of the 9th Workshop and 7th Workshop on Parallel Programming and RunTime Management Techniques for Manycore Architectures and Design Tools and Architectures for Multicore Embedded Computing PlatformsSince the introduction of Single Instruction Multiple Thread (SIMT) GPU architectures, vectorization has seldom been recommended. However, for efficient use of 8-bit and 16-bit data types, vector types are necessary even on these GPUs. When only integer ...
Directive-based parallelization of the NIM weather model for GPUs
WACCPD '14: Proceedings of the First Workshop on Accelerator Programming using DirectivesThe NIM is a performance-portable model that runs on CPU, GPU and MIC architectures with a single source code. The single source plus efficient code design allows application scientists to maintain the Fortran code, while computer scientists optimize ...
Numerical weather model BRAMS evaluation on many-core architectures: a micro and macro vision
This paper investigates the performance of a weather forecasting application Brazilian developments on the regional atmospheric modelling system - BRAMS on high performance computing HPC clusters with a multi-core architecture. We simulated atmosphere ...
Comments