Toward exascale whole-device modeling of fusion devices: Porting the GENE gyrokinetic microturbulence code to GPU

Germaschewski, K.; Allen, Bryce; Dannert, T.; Hrywniak, Markus; Donaghy, J. J.; Merlo, Gabriele; Ethier, S.; D'Azevedo, Eduardo F.; Jenko, F.; Bhattacharjee, A.

doi:10.1063/5.0046327

Cited by 24 publications

(6 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With simulation data written to file at every 150 Ω −1 i0 the total file size of one simulation is about 500 GB. The grid size is of similar order of magnitude as other fluid-type simulation runs [35,44] but is a factor 10 larger than the spatial grid size of some of the largest (computationally) gyro-kinetic runs [69,70]. We assume that a gyro-fluid model does not differ much in the spatial resolution requirement from its underlying gyro-kinetic model.…”

Section: Performance Observationsmentioning

confidence: 99%

Effects of plasma resistivity in FELTOR simulations of three-dimensional full-F gyro-fluid turbulence

Wiesenberger,

Held

2024

Plasma Phys. Control. Fusion

View full text Add to dashboard Cite

A full-F, isothermal, electromagnetic, gyro-fluid model is used to simulate plasma turbulence in a COMPASS-sized, diverted tokamak. A parameter scan covering three orders of magnitude of plasma resistivityand two values for the ion to electron temperature ratio with otherwise fixed parameters is setup and analysed.Two transport regimes for high and low plasma resistivities are revealed. Beyond a critical resistivitythe mass and energy confinement reduces with increasing resistivity. Further, for high plasma resistivity the direction of parallel acceleration is swapped compared to low resistivity. Three-dimensional visualisations using ray tracing techniques are displayed and discussed. The field-alignment of turbulent fluctuations in density andparallel current becomes evident.Relative density fluctuation amplitudes increase from below $1\%$ in the core to $15\%$ in the edge and up to $40\%$ in the scrape-off layer. Finally, the integration of exact conservation laws over the closed field line region allows for an identification of numerical errors within the simulations. The electron force balance and energy conservation show relative errors on the order of $10^{-3}$ while the particle conservation and ion momentum balance show errors on the order of $10^{-2}$.  All simulations are performed with a new version of the FELTOR code, which is fully parallelized on GPUs. Each simulation covers a couple of milliseconds of turbulence.

show abstract

Section: Performance Observationsmentioning

confidence: 99%

Effects of plasma resistivity in FELTOR simulations of three-dimensional full-F gyro-fluid turbulence

Wiesenberger,

Held

2024

Plasma Phys. Control. Fusion

View full text Add to dashboard Cite

show abstract

“…Finally, in the parallel velocity dimension, we adopt a box size of l v k ¼ 2 ffiffi ffi 2 p v i , while in the magnetic moment dimension, we utilize a box size of l l ¼ 3 T i =B 0 . The simulations were conducted using the GPU version of GENE 44 on the GPU partition of the Marconi supercomputer (Marconi100) at CINECA. The computational resources allocated for the simulations consisted of two nodes, with each node running for a duration of 2 min per iteration.…”

Section: Simulation Setupmentioning

confidence: 99%

Assessing core ion thermal confinement in critical-gradient-optimized stellarators

Bañón Navarro,

Roberg-Clark,

Plunk

et al. 2024

Physics of Plasmas

Self Cite

View full text Add to dashboard Cite

We investigate the core confinement properties of two recently devised quasi-helically symmetric stellarator configurations, HSK and QSTK. Both have been optimized for large critical gradients of the ion temperature gradient mode, which is an important driver of turbulent transport in magnetic confinement fusion devices. To predict the resulting core plasma profiles, assuming a fixed edge temperature, we utilize an advanced theoretical framework based on the gyrokinetic codes GENE and GENE-3D, coupled to the transport code TANGO. Compared to the HSX stellarator, both HSK and QSTK achieve significantly higher core-to-edge temperature ratios, partly thanks to their smaller aspect ratios, with the other part due to more detailed shaping of the magnetic geometry achieved during optimization. The computed core confinement time, however, is less sensitive to core temperature than the fixed edge temperature, simply due to the disproportionate influence, the edge has on stored plasma energy. We, therefore, emphasize the possible benefits of further optimizing turbulence in the outer core region, and the need to include accurate modeling of confinement in the edge region in order to assess overall plasma performance of turbulence optimized stellarators.

show abstract

“…Most of the existing work has been carried out on reduced setups (adiabatic electrons, collisionless and/or electrostatic plasma, simplified geometry, no impurities nor supra-thermal particles) [49][50][51][52]. In recent years, an increasing effort has been spent on extending gyrokinetic codes improving code performances, e.g., via GPU porting [53]; and developing new algorithms exploiting the time separation between micro-and macrophysics, e.g., time-telescoping methods [32,34,54]. One of the most promising approaches developed so far allowing gyrokinetic simulations up to the transport timescale, consists of coupling the gyrokinetic code to a 1D transport code (multiple-timescale approach).…”

Section: Theoretical Frameworkmentioning

confidence: 99%

Global gyrokinetic simulations of ASDEX Upgrade up to the transport timescale with GENE–Tango

Siena¹,

Navarro²,

Luda³

et al. 2022

Nucl. Fusion

Self Cite

View full text Add to dashboard Cite

An accurate description of turbulence up to the transport time scale is essential for predicting core plasma profiles and enabling reliable calculations for designing advanced scenarios and future devices. Here, we exploit the gap separation between turbulence and transport time scales and couple the global gyrokinetic code GENE to the transport-solver Tango, including kinetic electrons, collisions, realistic geometries, toroidal rotation and electromagnetic effects for the first time. This approach overcomes gyrokinetic codes' limitations and enables high-fidelity profile calculations in experimentally relevant plasma conditions, significantly reducing the computational cost. We present numerical results of GENE-Tango for two ASDEX Upgrade discharges, one of which exhibits a pronounced peaking of the ion temperature profile not reproduced by TGLF-ASTRA. We show that GENE-Tango can correctly capture the ion temperature peaking observed in the experiment. By retaining different physical effects in the GENE simulations, e.g., collisions, toroidal rotation and electromagnetic effects, we demonstrate that the ion temperature profile's peaking is due to electromagnetic effects of submarginal (stable) KBM modes. Based on these results, the expected GENE-Tango speedup for the ITER standard scenario is larger than two orders of magnitude compared to a single gyrokinetic simulation up to the transport time scale, possibly making first-principles ITER simulations feasible on current computing resources.

show abstract

Toward exascale whole-device modeling of fusion devices: Porting the GENE gyrokinetic microturbulence code to GPU

Cited by 24 publications

References 32 publications

Effects of plasma resistivity in FELTOR simulations of three-dimensional full-F gyro-fluid turbulence

Effects of plasma resistivity in FELTOR simulations of three-dimensional full-F gyro-fluid turbulence

Assessing core ion thermal confinement in critical-gradient-optimized stellarators

Global gyrokinetic simulations of ASDEX Upgrade up to the transport timescale with GENE–Tango

Contact Info

Product

Resources

About