Abstract of TRINCODE Project

Trinity is a Fortran transport code tightly integrated with the gyrokinetic flux-tube code GS2. Trinity-GS2 takes advantage of the separation in both space and time between turbulent and transport scales, and is used to support the medium-sized tokamak, JET and ITER physics programmes by performing first-principles simulations of turbulent transport across the plasma. Understanding the effect of plasma turbulence on confinement is fundamental to the early realization of fusion as an energy source.

Within the running HLST project TRINGS2, Trinity and GS2 have been modified to enable different numbers of cores to be used by different fluxtube simulations, thereby improving load-balancing by allowing available resources to be deployed where most needed. Separately, Trinity has recently been integrated with the C, CUDA-enabled gyrofluid code GryffX, which takes advantage of the GPU resources available on some platforms. Integration with the precursor Fortran gyrofluid code Gryffin is also under way. Gyrofluid models are approximations to the full gyrokinetic physics but can require orders of magnitude fewer computational resources for a given problem.

Within the present proposed project TRINCODE we aim to complete the Trinity-Gryffin integration, thereby allowing extensive exploration of parameter space on non-GPU HPC platforms such as Helios. We propose to develop an automated framework for running multiple instances of Trinity with different parameters including equilibrium magnetic geometries. This will expedite research programmes such as the predictive study of JET hybrid shots envisaged as part of the 3rd-cyle Helios project GKMSC. We also propose to increase the flexibility of Trinity (taking advantage of the changes already implemented in TRINGS2) by enabling it to use different flux codes or models at different plasma radii. This will further improve resource allocation and allow detailed studies to be focused on particular regions of interest.