Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Bücher
Zeitschriften
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
SLX-Digitalkonferenz: Zukunftswerkstatt 2024

Mehr Umsatz mit Top-Kontakten

Am 7. Mai erfahren Sie, wie Vertriebsteams Leadgenerierung, Leadnurturing und Leadstrategien effizient steuern können.
Erweiterte Suche
Anmelden
nach oben
Erschienen in:
An Example of Porting PETSc Applications to Heterogeneous Platforms with OpenACC Kapitel lesen Erstes Kapitel lesen

2018 | OriginalPaper | Buchkapitel

Hybrid Fortran: High Productivity GPU Porting Framework Applied to Japanese Weather Prediction Model

verfasst von : Michel Müller, Takayuki Aoki

Erschienen in: Accelerator Programming Using Directives

Verlag: Springer International Publishing

Einloggen

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config
loading …

Abstract

In this work we use the GPU porting task for the operative Japanese weather prediction model “ASUCA” as an opportunity to examine productivity issues with OpenACC when applied to structured grid problems. We then propose “Hybrid Fortran”, an approach that combines the advantages of directive based methods (no rewrite of existing code necessary) with that of stencil DSLs (memory layout is abstracted). This gives the ability to define multiple parallelizations with different granularities in the same code. Without compromising on performance, this approach enables a major reduction in the code changes required to achieve a hybrid GPU/CPU parallelization - as demonstrated with our ASUCA implementation using Hybrid Fortran.

Sie haben noch keine Lizenz? Dann Informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

Jetzt informieren

Springer Professional "Technik"

Online-Abonnement

Mit Springer Professional "Technik" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 390 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Maschinenbau + Werkstoffe




 

Jetzt Wissensvorsprung sichern!

Jetzt informieren

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 340 Zeitschriften

aus folgenden Fachgebieten:

  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Versicherung + Risiko




Jetzt Wissensvorsprung sichern!

Jetzt informieren
Vorheriges Kapitel An Example of Porting PETSc Applications to Heterogeneous Platforms with OpenACC
Nächstes Kapitel Implicit Low-Order Unstructured Finite-Element Multiple Simulation Enhanced by Dense Computation Using OpenACC
Fußnoten
1
These open-sourced efforts can be found at https://​github.​com/​C2SM-RCM/​claw-compiler.
 
3
Privatization is the main difference: Hybrid Fortran generated OpenMP code uses “firstprivate” as the default policy with an explicit “shared” clause for all arrays used in the kernel.
 
4
OpenACC is mainly used for reduction support - Hybrid Fortran does not automatically generate reduction kernels, however it supports the “reduce” clause, which is forwarded to the generated OpenMP or OpenACC kernels.
 
5
Thus obviating the need for code duplication and/or deep inlining of call trees.
 
6
Please note: While this paper follows American English, the Hybrid Fortran language extension has originally been developed following British English, which becomes apparent in the spelling of “domainDependant” (https://​en.​oxforddictionari​es.​com/​definition/​dependant). We consider support for the American English spelling of this directive as part of our future efforts.
 
8
Reduction kernels are thus not supported with this backend - we use the OpenACC backend selectively for this purpose, see also the discussion in the footnotes to Sect. 2.1.
 
10
Hybrid Fortran allows the user to switch between varying backend implementations per routine, such as OpenACC and CUDA Fortran - the user specified information as well as the defaults given by the build system call thus steers this implementation class.
 
11
Alternatively, a dependency generator script can be configured as well.
 
12
Since the input to this analysis is the closed source ASUCA codebase, full reproducibility cannot be provided in this context. However the intermediate data, the method employed to gather this data as well as a sample input is provided and documented in https://​github.​com/​muellermichel/​hybrid-asuca-productivity-evidence/​blob/​master/​asuca_​productivity.​xlsx.
 
13
Please refer to https://​github.​com/​muellermichel/​Hybrid-Fortran/​blob/​v1.​00rc10/​examples/​Overview.​md for an overview of the available samples and their results.
 
Literatur
1.
Cumming, B., Osuna, C., Gysi, T., Bianco, M., Lapillonne, X., Fuhrer, O., Schulthess, T.C.: A review of the challenges and results of refactoring the community climate code COSMO for hybrid Cray HPC systems. In: Proceedings of Cray User Group (2013)
2.
Douglas, C.C., Hu, J., Kowarschik, M., Rüde, U., Weiß, C.: Cache optimization for structured and unstructured grid multigrid. Electron. Trans. Numer. Anal. 10, 21–40 (2000)MathSciNetMATH
3.
Dursun, H., Nomura, K.I., Wang, W., Kunaseth, M., Peng, L., Seymour, R., Kalia, R.K., Nakano, A., Vashishta, P.: In-core optimization of high-order stencil computations. In: PDPTA, pp. 533–538 (2009)
4.
Edwards, H.C., Trott, C.R., Sunderland, D.: Kokkos: enabling manycore performance portability through polymorphic memory access patterns. J. Parallel Distrib. Comput. 74(12), 3202–3216 (2014). Domain-specific languages and high-level frameworks for high-performance computingCrossRef
5.
6.
Fuhrer, O., Osuna, C., Lapillonne, X., Gysi, T., Cumming, B., Bianco, M., Arteaga, A., Schulthess, T.C.: Towards a performance portable, architecture agnostic implementation strategy for weather and climate models. Supercomputing Front. Innovations 1(1), 45–62 (2014)
7.
Govett, M., Middlecoff, J., Henderson, T.: Directive-based parallelization of the NIM weather model for GPUs. In: 2014 First Workshop on Accelerator Programming using Directives (WACCPD), pp. 55–61. IEEE (2014)
8.
Govett, M., Rosinski, J., Middlecoff, J., Henderson, T., Lee, J., MacDonald, A., Wang, N., Madden, P., Schramm, J., Duarte, A.: Parallelization and performance of the NIM weather model on CPU, GPU and MIC processors. Bulletin of the American Meteorological Society (2017)
9.
Gysi, T., Hoefler, T.: Integrating STELLA & MODESTO: definition and optimization of complex stencil programs (2017)
10.
Ishida, J., Muroi, C., Kawano, K., Kitamura, Y.: Development of a new nonhydrostatic model ASUCA at JMA. CAS/JSC WGNE Res. Activities Atmos. Oceanic Model. 40, 0511–0512 (2010)
11.
Jumah, N., Kunkel, J., Zängl, G., Yashiro, H., Dubos, T., Meurdesoif, Y.: GGDML: icosahedral models language extensions (2017)
12.
13.
Lapillonne, X., Fuhrer, O.: Using compiler directives to port large scientific applications to GPUs: an example from atmospheric science. Parallel Process. Lett. 24(01), 1450003 (2014)MathSciNetCrossRef
14.
Mielikainen, J., Huang, B., Huang, A.: Using Intel Xeon Phi to accelerate the WRF TEMF planetary boundary layer scheme. In: SPIE Sensing Technology + Applications, p. 91240T. International Society for Optics and Photonics (2014)
15.
Müller, M., Aoki, T.: New high performance GPGPU code transformation framework applied to large production weather prediction code (2017, to be published in ACM TOPC)
16.
Norman, M.R., Mametjanov, A., Taylor, M.: Exascale programming approaches for the accelerated model for climate and energy (2017)
17.
Quinlan, D.: ROSE: compiler support for object-oriented frameworks. Parallel Process. Lett. 10(02n03), 215–226 (2000)CrossRef
18.
Sakamoto, M., Ishida, J., Kawano, K., Matsubayashi, K., Aranami, K., Hara, T., Kusabiraki, H., Muroi, C., Kitamura, Y.: Development of yin-yang grid global model using a new dynamical core ASUCA (2014)
19.
Sawyer, W., Zaengl, G., Linardakis, L.: Towards a multi-node OpenACC implementation of the ICON model. In: EGU General Assembly Conference Abstracts, vol. 16 (2014)
20.
Shimokawabe, T., Aoki, T., Muroi, C., Ishida, J., Kawano, K., Endo, T., Nukada, A., Maruyama, N., Matsuoka, S.: An 80-fold speedup, 15.0 TFlops full GPU acceleration of non-hydrostatic weather model ASUCA production code. In: Proceedings of the 2010 ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis, pp. 1–11. IEEE Computer Society (2010)
21.
Shimokawabe, T., Aoki, T., Onodera, N.: High-productivity framework on GPU-rich supercomputers for operational weather prediction code ASUCA. In: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2014, pp. 251–261. IEEE Press, Piscataway (2014). https://​doi.​org/​10.​1109/​SC.​2014.​26
22.
Torres, R., Linardakis, L., Kunkel, J., Ludwig, T.: ICON DSL: A domain-specific language for climate modeling. In: International Conference for High Performance Computing, Networking, Storage and Analysis, Denver, CO (2013)
23.
Wahib, M., Maruyama, N.: Scalable kernel fusion for memory-bound GPU applications. In: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2014, pp. 191–202. IEEE Press, Piscataway (2014). https://​doi.​org/​10.​1109/​SC.​2014.​21
24.
Wicker, L.J., Skamarock, W.C.: Time-splitting methods for elastic models using forward time schemes. Mon. Weather Rev. 130(8), 2088–2097 (2002)CrossRef
Metadaten
Titel
Hybrid Fortran: High Productivity GPU Porting Framework Applied to Japanese Weather Prediction Model
verfasst von
Michel Müller
Takayuki Aoki
Copyright-Jahr
2018
Buch
Accelerator Programming Using Directives

Print ISBN: 978-3-319-74895-5

Electronic ISBN: 978-3-319-74896-2

Copyright-Jahr: 2018

DOI
https://doi.org/10.1007/978-3-319-74896-2_2