Simulation data provenance: "GPU-enabled Extreme Scale Turbulence Simulations (GESTS)" code at Georgia Tech
See README-isotropic32768 for more details✪ Direct numerical simulation (DNS) using 32,7683 grid-points.
✪ Code was run on AMD MI-250X GPUs on the world's first exascale computer (Frontier, at US DOE Oak Ridge National Laboratory).
✪ Navier-Stokes is solved using pseudo-spectral method.
✪ Time integration uses second-order Runge-Kutta.
✪ The simulation is de-aliased using phase-shifting and truncation.
✪ Energy is injected by keeping the energy density in the lowest wavenumber modes prescribed following the approach of Donzis & Yeung.
✪ The simulation is run by reducing the viscosity and refining the grid in steps, which enables reaching near equilibrated conditions after a few respective small-scale turn-over times.
✪ After the simulation has reached a statistical quasi-stationary state, a frame of data, which includes the 3 components of the velocity vector and the pressure, is written in files that can be accessed directly by the database (Zarr file on ceph cluster).
✪ Domain: 2π × 2π × 2π
✪ Grid: 32,7683
✪ Number of snapshots available: 1
✪ Taylor-scale Reynolds number Reλ~ 2,500
✪ Viscosity, dissipation, RMS velocity, and Kolmogorov scale: see README-isotropic32768
✪ GetPosition is not implemented for this dataset.