GPU-optimized fast plasma equilibrium reconstruction in fine grids for real-time control and data analysis

Abstract

P-EFIT, a GPU parallel equilibrium reconstruction code, is based on the EFIT framework, but built with the CUDA™ (Compute Unified Device Architecture) to take advantage of massively parallel Graphical Processing Unit (GPU) cores to significantly accelerate the computation. With the parallelized Grad-Shafranov solver and middle-scale matrix calculation modules, P-EFIT can accurately reproduce the EFIT reconstruction algorithms at a fraction of the EFIT computational time. Integrated into EAST plasma control system, P-EFIT not only provides control signal results but also enhances EAST plasma control capacity with unique designed function modules. Using the synthetic magnetic diagnostic signals from ITER plasma equilibria obtained by the CREATE-L and CREATE-NL codes in standalone and streaming mode, P-EFIT has good enough accuracy and time latency performance in most of the considered cases. P-EFIT achieves full kinetic equilibrium reconstruction algorithms and repeats the EFIT results with the DIII-D internal plasma current and kinetic profile measurements in one-percent cost time of EFIT. All these works suggest that P-EFIT can provide quality magnetic equilibrium reconstruction in real-time, offer full kinetic equilibrium reconstruction with high spatial resolution and high speed, support more detailed, better plasma control and data analysis for tokamak devices.