Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware

A reduction of computational cost would allow higher resolution in numerical weather predictions within the same budget for computation. This paper investigates two approaches that promise significant savings in computational cost: the use of reduced precision hardware, which reduces floating point...

Popoln opis

Bibliografske podrobnosti
Main Authors: Dueben, P, Palmer, T
Format: Journal article
Jezik:English
Izdano: American Meteorological Society 2014
_version_ 1826299474443501568
author Dueben, P
Palmer, T
author_facet Dueben, P
Palmer, T
author_sort Dueben, P
collection OXFORD
description A reduction of computational cost would allow higher resolution in numerical weather predictions within the same budget for computation. This paper investigates two approaches that promise significant savings in computational cost: the use of reduced precision hardware, which reduces floating point precision beyond the standard double- and single-precision arithmetic, and the use of stochastic processors, which allow hardware faults in a trade-off between reduced precision and savings in power consumption and computing time. Reduced precision is emulated within simulations of a spectral dynamical core of a global atmosphere model and a detailed study of the sensitivity of different parts of the model to inexact hardware is performed. Afterward, benchmark simulations were performed for which as many parts of the model as possible were put onto inexact hardware. Results show that large parts of the model could be integrated with inexact hardware at error rates that are surprisingly high or with reduced precision to only a couple of bits in the significand of floating point numbers. However, the sensitivities to inexact hardware of different parts of the model need to be respected, for example, via scale separation. In the last part of the paper, simulations with a full operational weather forecast model in single precision are presented. It is shown that differences in accuracy between the single- and double-precision forecasts are smaller than differences between ensemble members of the ensemble forecast at the resolution of the standard ensemble forecasting system. The simulations prove that the trade-off between precision and performance is a worthwhile effort, already on existing hardware.
first_indexed 2024-03-07T05:02:28Z
format Journal article
id oxford-uuid:d8c3caf3-5c8a-4acf-80e1-bc36d0c69cb8
institution University of Oxford
language English
last_indexed 2024-03-07T05:02:28Z
publishDate 2014
publisher American Meteorological Society
record_format dspace
spelling oxford-uuid:d8c3caf3-5c8a-4acf-80e1-bc36d0c69cb82022-03-27T08:51:07ZBenchmark Tests for Numerical Weather Forecasts on Inexact HardwareJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:d8c3caf3-5c8a-4acf-80e1-bc36d0c69cb8EnglishSymplectic Elements at OxfordAmerican Meteorological Society2014Dueben, PPalmer, TA reduction of computational cost would allow higher resolution in numerical weather predictions within the same budget for computation. This paper investigates two approaches that promise significant savings in computational cost: the use of reduced precision hardware, which reduces floating point precision beyond the standard double- and single-precision arithmetic, and the use of stochastic processors, which allow hardware faults in a trade-off between reduced precision and savings in power consumption and computing time. Reduced precision is emulated within simulations of a spectral dynamical core of a global atmosphere model and a detailed study of the sensitivity of different parts of the model to inexact hardware is performed. Afterward, benchmark simulations were performed for which as many parts of the model as possible were put onto inexact hardware. Results show that large parts of the model could be integrated with inexact hardware at error rates that are surprisingly high or with reduced precision to only a couple of bits in the significand of floating point numbers. However, the sensitivities to inexact hardware of different parts of the model need to be respected, for example, via scale separation. In the last part of the paper, simulations with a full operational weather forecast model in single precision are presented. It is shown that differences in accuracy between the single- and double-precision forecasts are smaller than differences between ensemble members of the ensemble forecast at the resolution of the standard ensemble forecasting system. The simulations prove that the trade-off between precision and performance is a worthwhile effort, already on existing hardware.
spellingShingle Dueben, P
Palmer, T
Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title_full Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title_fullStr Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title_full_unstemmed Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title_short Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware
title_sort benchmark tests for numerical weather forecasts on inexact hardware
work_keys_str_mv AT duebenp benchmarktestsfornumericalweatherforecastsoninexacthardware
AT palmert benchmarktestsfornumericalweatherforecastsoninexacthardware