Peter D. Düben

On the use of inexact, pruned hardware in atmospheric modelling

Inexact hardware design, which advocates trading the accuracy of computations in exchange for significant savings in area, power and/or performance of computing hardware, has received increasing prominence in several error-tolerant application domains, particularly those involving perceptual or statistical end-users. In this paper, we evaluate inexact hardware for its applicability in weather and climate modelling. We expand previous studies on inexact techniques, in particular probabilistic pruning, to floating point arithmetic units and derive several simulated set-ups of pruned hardware with reasonable levels of error for applications in atmospheric modelling. The set-up is tested on the Lorenz ‘96 model, a toy model for atmospheric dynamics, using software emulation for the proposed hardware. The results show that large parts of the computation tolerate the use of pruned hardware blocks without major changes in the quality of short- and long-time diagnostics, such as forecast errors and probability density functions. This could open the door to significant savings in computational cost and to higher resolution simulations with weather and climate models.

Benchmark Tests for Numerical Weather Forecasts on Inexact Hardware

AbstractA 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 redu...

Opportunities for energy efficient computing: a study of inexact general purpose processors for high-performance and big-data applications

In this paper, we demonstrate that disproportionate gains are possible through a simple devise for injecting inexactness or approximation into the hardware architecture of a computing system with a general purpose template including a complete memory hierarchy. The focus of the study is on energy savings possible through this approach in the context of large and challenging applications. We choose two such from different ends of the computing spectrum-the IGCM model for weather and climate modeling which embodies significant features of a high-performance computing workload, and the ubiquitous PageRank algorithm used in Internet search. In both cases, we are able to show in the affirmative that an inexact system outperforms its exact counterpart in terms of its efficiency quantified through the relative metric of operations per virtual Joule (OPVJ)-a relative metric that is not tied to particular hardware technology. As one example, the IGCM application can be used to achieve savings through inexactness of (almost) a factor of 3 in energy without compromising the quality of the forecast, quantified through the forecast error metric, in a noticeable manner. As another example finding, we show that in the case of PageRank, an inexact system is able to outperform its exact counterpart by close to a factor of 1.5 using the OPVJ metric.

A discontinuous/continuous low order finite element shallow water model on the sphere

We study the applicability of a new finite element in atmosphere and ocean modeling. The finite element under investigation combines a second order continuous representation for the scalar field with a first order discontinuous representation for the velocity field and is therefore different from continuous and discontinuous Galerkin finite element approaches. The specific choice of low order approximation spaces is attractive because it satisfies the Ladyzhenskaya-Babuska-Brezzi condition and is, at the same time, able to represent the crucially important geostrophic balance. The finite element is used to solve the viscous and inviscid shallow water equations on a rotating sphere. We introduce the spherical geometry via a stereographic projection. The projection leads to a manageable number of additional terms, the associated scaling factors can be exactly represented by second order polynomials. We perform numerical experiments considering steady and unsteady zonal flow, flow over topography, and an unstable zonal jet stream. For ocean applications, the wind driven Stommel gyre is simulated. The experiments are performed on icosahedral geodesic grids and analyzed with respect to convergence rates, conservation properties, and energy and enstrophy spectra. The results match quite well with results published in the literature and encourage further investigation of this type of element for three-dimensional atmosphere/ocean modeling.

Single Precision in Weather Forecasting Models: An Evaluation with the IFS

AbstractEarth’s climate is a nonlinear dynamical system with scale-dependent Lyapunov exponents. As such, an important theoretical question for modeling weather and climate is how much real information is carried in a model’s physical variables as a function of scale and variable type. Answering this question is of crucial practical importance given that the development of weather and climate models is strongly constrained by available supercomputer power. As a starting point for answering this question, the impact of limiting almost all real-number variables in the forecasting mode of ECMWF Integrated Forecast System (IFS) from 64 to 32 bits is investigated. Results for annual integrations and medium-range ensemble forecasts indicate no noticeable reduction in accuracy, and an average gain in computational efficiency by approximately 40%. This study provides the motivation for more scale-selective reductions in numerical precision.

Architectures and Precision Analysis for Modelling Atmospheric Variables with Chaotic Behaviour

The computationally intensive nature of atmospheric modelling is an ideal target for hardware acceleration. Performance of hardware designs can be improved through the use of reduced precision arithmetic, but maintaining appropriate accuracy is essential. We explore reduced precision optimisation for simulating chaotic systems, targeting atmospheric modelling in which even minor changes in arithmetic behaviour can have a significant impact on system behaviour. Hence, standard techniques for comparing numerical accuracy are inappropriate. We use the Hellinger distance to compare statistical behaviour between reduced-precision CPU implementations to guide FPGA designs of a chaotic system, and analyse accuracy, performance and power efficiency of the resulting implementations. Our results show that with only a limited loss in accuracy corresponding to less than 10% uncertainly in input parameters, a single Xilinx Virtex 6 SXT475 FPGA can be 13 times faster and 23 times more power efficient than a 6-core Intel Xeon X5650 processor.

Stochastic modelling and energy-efficient computing for weather and climate prediction

Being able to predict weather and climate reliably over a range of time scale is crucial if we are to create a society which is resilient to extremes of weather and climate. Although current weather and climate models can show impressive levels of predictive skill, for example in predicting the

A study of reduced numerical precision to make superparameterization more competitive using a hardware emulator in the OpenIFS model

The use of reduced numerical precision to reduce computing costs for the cloud resolving model of superparameterized simulations of the atmosphere is investigated. An approach to identify the optimal level of precision for many different model components is presented, and a detailed analysis of precision is performed. This is nontrivial for a complex model that shows chaotic behavior such as the cloud resolving model in this paper. It is shown not only that numerical precision can be reduced significantly but also that the results of the reduced precision analysis provide valuable information for the quantification of model uncertainty for individual model components. The precision analysis is also used to identify model parts that are of less importance thus enabling a reduction of model complexity. It is shown that the precision analysis can be used to improve model efficiency for both simulations in double precision and in reduced precision. Model simulations are performed with a superparameterized single-column model version of the OpenIFS model that is forced by observational data sets. A software emulator was used to mimic the use of reduced precision floating point arithmetic in simulations.

Lower precision for higher accuracy: Precision and resolution exploration for shallow water equations

Accurate forecasts of future climate with numerical models of atmosphere and ocean are of vital importance. However, forecast quality is often limited by the available computational power. This paper investigates the acceleration of a C-grid shallow water model through the use of reduced precision targeting FPGA technology. Using a double-gyre scenario, we show that the mantissa length of variables can be reduced to 14 bits without affecting the accuracy beyond the error inherent in the model. Our reduced precision FPGA implementation runs 5.4 times faster than a double precision FPGA implementation, and 12 times faster than a multi-threaded CPU implementation. Moreover, our reduced precision FPGA implementation uses 39 times less energy than the CPU implementation and can compute a 100×100 grid for the same energy that the CPU implementation would take for a 29×29 grid.

Monte Carlo simulations of the randomly forced Burgers equation

The behaviour of the one-dimensional random-forced Burgers equation is investigated in the path integral formalism, using a discrete space-time lattice. We show that by means of Monte Carlo methods one may evaluate observables, such as structure functions, as ensemble averages over different field realizations. The regularization of shock solutions to the zero-viscosity limit (Hopf equation) eventually leads to constraints on lattice parameters, required for the stability of the simulations. Insight into the formation of localized structures (shocks) and their dynamics is obtained.