Abstract
EFIT is one of the most extensively used equilibrium reconstruction code in the world. Although robust, EFIT plasma reconstructions will face new challenges in the burning plasma era. These include adapting to new operating regimes and relying on diagnostics that can survive in a harsher, radioactive environment. These stricter plasma control scenarios have motivated exploration of machine learning techniques to improve the quality of real-time equilibrium reconstructions. In order to train these new methods, a database of solutions is required which carefully tracks all constraints and fits performed by EFIT. To support these new developments, we are also improving the core Grad-Shafranov solver. Changes include clearly separating device-specific coding, improving code portability, developing a continuous development pipeline with automatic regression testing, and ensuring thread-safety in preparation for GPU-developments. Using extremely-portable OpenMP and OpenACC directives, we have been able to improve the performance of EFIT using GPU hardware. For the subroutines tested we have observed 40 times speedup across multiple GPU vendors. New options have also been added to assist with creation and analysis of database results. Access to these new techniques is made widely available with Gitlab hosted documentation [https://efit-ai.gitlab.io/efit/] and integration with the OMFIT framework [https://omfit.io] and existing workflows.
*This material is based upon work supported by the Department of Energy under Award Number(s) DE-SC0021203. Disclaimer: This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.
