On the use of inexact, pruned hardware in atmospheric modelling

Dueben, Peter; Joven, Jorge; Lingamneni, Avinash; McNamara, Hugh; Micheli, Giovanni De; Palem, Krishna V.; Palmer, Tim

doi:10.1098/rsta.2013.0276

Cited by 34 publications

(33 citation statements)

References 21 publications

(50 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While individual components show some speedup, whole MPMD programs show only modest increases in performance (see e.g., Govett et al, 2014;Iacono et al, 2014;Fuhrer et al, 2014;Ford et al, 2014). Other approaches, such as the use of inexact computing (Korkmaz et al, 2006;Düben et al, 2014), are still in very early stages.…”

Section: Discussionmentioning

confidence: 99%

Coarse-grained component concurrency in Earth system modeling: parallelizing atmospheric radiative transfer in the GFDL AM3 model using the Flexible Modeling System coupling framework

et al. 2016

View full text Add to dashboard Cite

Abstract. Climate models represent a large variety of processes on a variety of timescales and space scales, a canonical example of multi-physics multi-scale modeling. Current hardware trends, such as Graphical Processing Units (GPUs) and Many Integrated Core (MIC) chips, are based on, at best, marginal increases in clock speed, coupled with vast increases in concurrency, particularly at the fine grain. Multiphysics codes face particular challenges in achieving finegrained concurrency, as different physics and dynamics components have different computational profiles, and universal solutions are hard to come by.We propose here one approach for multi-physics codes. These codes are typically structured as components interacting via software frameworks. The component structure of a typical Earth system model consists of a hierarchical and recursive tree of components, each representing a different climate process or dynamical system. This recursive structure generally encompasses a modest level of concurrency at the highest level (e.g., atmosphere and ocean on different processor sets) with serial organization underneath.We propose to extend concurrency much further by running more and more lower-and higher-level components in parallel with each other. Each component can further be parallelized on the fine grain, potentially offering a major increase in the scalability of Earth system models.We present here first results from this approach, called coarse-grained component concurrency, or CCC. Within the Geophysical Fluid Dynamics Laboratory (GFDL) Flexible Modeling System (FMS), the atmospheric radiative transfer component has been configured to run in parallel with a composite component consisting of every other atmospheric component, including the atmospheric dynamics and all other atmospheric physics components. We will explore the algorithmic challenges involved in such an approach, and present results from such simulations. Plans to achieve even greater levels of coarse-grained concurrency by extending this approach within other components, such as the ocean, will be discussed.

show abstract

Section: Discussionmentioning

confidence: 99%

Coarse-grained component concurrency in Earth system modeling: parallelizing atmospheric radiative transfer in the GFDL AM3 model using the Flexible Modeling System coupling framework

et al. 2016

View full text Add to dashboard Cite

show abstract

“…If FPGAs are used, the speed-up will be approximately proportional to the ratio of the number of bits used. An inexact CPU setup with a pruned floating-point unit and inexact memory might even scale better than the ratio between the number of bits [7,10]. We conclude that the reduced precision setup will be more than three times faster compared to the standard double precision simulation (64/19 ≈ 3.4).…”

Section: Performance Analysismentioning

confidence: 75%

“…While the large potential for the use of inexact hardware in atmosphere models was already shown in previous studies [7][8][9], this is the first study to test the use of inexact hardware in a grid point instead of a spectral model. We show, in a very idealised setup, that rounding errors can improve numerical simulations in ensemble methods in the same way stochastic forcings of stochastic parametrisation schemes can do.…”

Section: Discussionmentioning

confidence: 99%

“…The first approach to inexact hardware was to use so-called stochastic processors that allow hardware errors to occur within numerical simulations while power consumption is reduced by reducing the applied voltage to the floatingpoint unit (see for example [16,20]). A second approach is "pruning" for which the physical size of the floating-point unit is reduced, by removing parts that are either hardly used or do not have a strong influence on significant bits in the results of floating-point operations [7,17]. Pruning of the floating-point unit can be combined with the use of inexact memory [10].…”

mentioning

confidence: 99%

See 1 more Smart Citation

Rounding errors may be beneficial for simulations of atmospheric flow: results from the forced 1D Burgers equation

Dueben

Dolaptchiev

2015

Theor. Comput. Fluid Dyn.

Self Cite

View full text Add to dashboard Cite

Inexact hardware can reduce computational cost, due to a reduced energy demand and an increase in performance, and can therefore allow higher-resolution simulations of the atmosphere within the same budget for computation. We investigate the use of emulated inexact hardware for a model of the randomly forced 1D Burgers equation with stochastic sub-grid-scale parametrisation. Results show that numerical precision can be reduced to only 12 bits in the significand of floating-point numbers-instead of 52 bits for double precisionwith no serious degradation in results for all diagnostics considered. Simulations that use inexact hardware on a grid with higher spatial resolution show results that are significantly better compared to simulations in double precision on a coarser grid at similar estimated computing cost. In the second half of the paper, we compare the forcing due to rounding errors to the stochastic forcing of the stochastic parametrisation scheme that is used to represent sub-grid-scale variability in the standard model setup. We argue that stochastic forcings of stochastic parametrisation schemes can provide a first guess for the upper limit of the magnitude of rounding errors of inexact hardware that can be tolerated by model simulations and suggest that rounding errors can be hidden in the distribution of the stochastic forcing. We present an idealised model setup that replaces the expensive stochastic forcing of the stochastic parametrisation scheme with an engineered rounding error forcing and provides results of similar quality. The engineered rounding error forcing can be used to create a forecast ensemble of similar spread compared to an ensemble based on the stochastic forcing. We conclude that rounding errors are not necessarily degrading the quality of model simulations. Instead, they can be beneficial for the representation of sub-grid-scale variability.

show abstract

“…Researchers have previously proposed multiple hardware architecture designs that improve the performance or energy consumption of processors by reducing precision or increasing the incidence of error [8,9,15,18,19,28,30,35,44,46,47]. Researchers have also proposed a variety of approximate DRAM and SRAM designs.…”

Section: Related Workmentioning

confidence: 99%

Approximate computation with outlier detection in Topaz

Achour

Rinard

2015

SIGPLAN Not.

View full text Add to dashboard Cite

We present Topaz, a new task-based language for computations that execute on approximate computing platforms that may occasionally produce arbitrarily inaccurate results. Topaz maps tasks onto the approximate hardware and integrates the generated results into the main computation. To prevent unacceptably inaccurate task results from corrupting the main computation, Topaz deploys a novel outlier detection mechanism that recognizes and precisely reexecutes outlier tasks. Outlier detection enables Topaz to work effectively with approximate hardware platforms that have complex fault characteristics, including platforms with bit pattern dependent faults (in which the presence of faults may depend on values stored in adjacent memory cells). Our experimental results show that, for our set of benchmark applications, outlier detection enables Topaz to deliver acceptably accurate results (less than 1% error) on our target approximate hardware platforms. Depending on the application and the hardware platform, the overall energy savings range from 5 to 13 percent. Without outlier detection, only one of the applications produces acceptably accurate results.

show abstract

On the use of inexact, pruned hardware in atmospheric modelling

Cited by 34 publications

References 21 publications

Coarse-grained component concurrency in Earth system modeling: parallelizing atmospheric radiative transfer in the GFDL AM3 model using the Flexible Modeling System coupling framework

Coarse-grained component concurrency in Earth system modeling: parallelizing atmospheric radiative transfer in the GFDL AM3 model using the Flexible Modeling System coupling framework

Rounding errors may be beneficial for simulations of atmospheric flow: results from the forced 1D Burgers equation

Approximate computation with outlier detection in Topaz

Contact Info

Product

Resources

About