Mixed-Precision for Linear Solvers in Global Geophysical Flows.

Semi-implicit (SI) time-stepping schemes for atmosphere and ocean models require elliptic solvers that work efficiently on modern supercomputers. This paper reports our study of the potential computational savings when using mixed precision arithmetic in the elliptic solvers. Precision levels as low as half (16 bits) are used and a detailed evaluation of the impact of reduced precision on the solver convergence and the solution quality is performed. This study is conducted in the context of a novel SI shallow-water model on the sphere, purposely designed to mimic numerical intricacies of modern all-scale weather and climate (W&C) models. The governing algorithm of the shallow-water model is based on the non-oscillatory MPDATA methods for geophysical flows, whereas the resulting elliptic problem employs a strongly preconditioned non-symmetric Krylov-subspace Generalized Conjugated-Residual (GCR) solver, proven in advanced atmospheric applications. The classical longitude/latitude grid is deliberately chosen to retain the stiffness of global W&C models. The analysis of the precision reduction is done on a software level, using an emulator, whereas the performance is measured on actual reduced precision hardware. The reduced-precision experiments are conducted for established dynamical-core test-cases, like the Rossby-Haurwitz wavenumber 4 and a zonal orographic flow. The study shows that selected key components of the elliptic solver, most prominently the preconditioning and the application of the linear operator, can be performed at the level of half precision. For these components, the use of half precision is found to yield a speed-up of a factor 4 compared to double precision for a wide range of problem sizes.

Heterogeneities of flow in stochastically generated porous media.

Heterogeneous flows are observed to result from variations in the geometry and topology of pore structures within stochastically generated three dimensional porous media. A stochastic procedure generates media comprising complex networks of connected pores. Inside each pore space, the Navier-Stokes equations are numerically integrated until steady state velocity and pressure fields are attained. The intricate pore structures exert spatially variable resistance on the fluid, and resulting velocity fields have a wide range of magnitudes and directions. Spatially nonuniform fluid fluxes are observed, resulting in principal pathways of flow through the media. In some realizations, up to 25% of the flux occurs in 5% of the pore space depending on porosity. The degree of heterogeneity in the flow is quantified over a range of porosities by tracking particle trajectories and calculating their attributes including tortuosity, length, and first passage time. A representative elementary volume is first computed so the dependence of particle based attributes on the size of the domain through which they are followed is minimal. High correlations between the dimensionless quantities of porosity and tortuosity are calculated and a logarithmic relationship is proposed. As the porosity of a medium increases the flow field becomes more uniform.

Coupling the dynamics of boundary layers and evolutionary dunes.

A theoretical formulation and corresponding numerical solutions are presented for fluid flow and sediment transport past evolutionary sand dunes. Time-dependent curvilinear coordinates are employed to fully couple flow aloft with the developing landform. The differential conservation law that defines shape of the lower boundary depends on details of local surface stress, thereby favoring the large eddy simulation of the boundary layer. To shrink the gap between the time scales characteristic of planetary boundary layer flows O(10(3)) s and sand dune evolution O(10(6)) s, a hypothetical "severe-wind scenario" is adopted with the saltation flux amplified up to 3 orders of magnitude. While the results are largely insensitive to the rescaling, the efficacy of computations is greatly improved. The flux-form partial differential equation for the interface profile--via saltation and sand avalanches--is formulated as an advection-diffusion equation, to facilitate discrete integrations. Numerical experiments verify the adopted theoretical framework by reproducing scaling results reported in the literature. The versatility of the approach is illustrated with evolution of a sandhole--an example of application likely never addressed in the literature, yet realizable in nature.