Accelerating high-fidelity gyrokinetic fusion simulations with iGENE
Abstract
We present the latest version of iGENE, a differentiable implementation of the local (flux-tube) gyrokinetic code GENE, now fully rewritten in JAX. iGENE features a multispecies, collisional, fully electromagnetic nonlinear gyrokinetic model while preserving end-to-end differentiability. The code is verified against GENE for linear ITG, TEM and KBM physics and for nonlinear turbulent fluxes with adiabatic and kinetic electrons. Forward-mode derivatives of the linear growth rate agree with finite differences for all five parameters tested, resolving the 30--50\% discrepancy in ŝ and ε reported for the earlier TensorFlow version. All collectives are native JAX operations, and the code strong-scales from 2 to 32 H100 GPUs with 86\% parallel efficiency. We then assess three routes to a shorter time to solution. Custom CUDA FFT kernels operating in bfloat16 are up to 3.97× faster than fp32 cuFFT on the transforms, and on a single H100 iGENE is 2.63× faster per time step than single-precision GENE. Multi-rate sub-cycling of the parallel dynamics yields a further 1.89× speedup with kinetic electrons. Finally, a Hermite-Laguerre velocity-space representation is implemented, offering tunable fidelity.
Related
- Minute-Scale High-Fidelity Gyrokinetic Simulations with Portability from Laptop to Supercomputer
- Design of the IMAS interface data structure for flux-tube gyrokinetic simulations
- Mitigation of Initial Transients in Total-f Gyrokinetic Turbulence Simulations Using Neoclassically Relaxed Distribution Function
- Global Gyrokinetic Simulations of Electromagnetic Turbulence in STEP
- Energetic particles, Shafranov shift and finite β effects on TAE, KBM and ITG instabilities in global electromagnetic gyrokinetic simulations
BibTeX
@misc{merlo2026accelerating,
title = {{Accelerating high-fidelity gyrokinetic fusion simulations with iGENE}},
author = {Gabriele Merlo and Victor Artigues and Carl-Martin Pfeiler and Frank Jenko},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.33561},
eprint = {2610.33561},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.33561},
}
RIS
TY - UNPB
TI - Accelerating high-fidelity gyrokinetic fusion simulations with iGENE
AU - Gabriele Merlo
AU - Victor Artigues
AU - Carl-Martin Pfeiler
AU - Frank Jenko
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2610.33561
UR - https://arxiv.org/abs/2610.33561
AB - We present the latest version of iGENE, a differentiable implementation of the local (flux-tube) gyrokinetic code GENE, now fully rewritten in JAX. iGENE features a multispecies, collisional, fully electromagnetic nonlinear gyrokinetic model while preserving end-to-end differentiability. The code is verified against GENE for linear ITG, TEM and KBM physics and for nonlinear turbulent fluxes with adiabatic and kinetic electrons. Forward-mode derivatives of the linear growth rate agree with finite differences for all five parameters tested, resolving the 30--50\% discrepancy in ŝ and ε reported for the earlier TensorFlow version. All collectives are native JAX operations, and the code strong-scales from 2 to 32 H100 GPUs with 86\% parallel efficiency. We then assess three routes to a shorter time to solution. Custom CUDA FFT kernels operating in bfloat16 are up to 3.97× faster than fp32 cuFFT on the transforms, and on a single H100 iGENE is 2.63× faster per time step than single-precision GENE. Multi-rate sub-cycling of the parallel dynamics yields a further 1.89× speedup with kinetic electrons. Finally, a Hermite-Laguerre velocity-space representation is implemented, offering tunable fidelity.
ER -