Performance Improvement in Real-Time Mapping of Thomson Scattering Data to Flux Coordinates in LHD

摘要

More than 100 diagnostic devices are attached to the vacuum vessel of the Large Helical Device (LHD); they measure various aspects of the plasma physics. Because the shape of the LHD plasma is not symmetric, each diagnostic obtains the physical values in a different cross section. For example, the Thomson scattering system measures the electron temperature profile in the horizontally elongated cross section, and the laser interferometer measures the line-integrated electron density profile in the vertically elongated cross section. To analyze the data obtained by different diagnostics, their measurement positions must be mapped to a unified coordinate system, the flux coordinate system. Therefore, the authors have been building a database to map the physical coordinates to the flux coordinates. A system for mapping the electron temperature profile to the flux coordinates, TSMAP, has been developed using the database. The profiles calculated by TSMAP are fundamental data for analyzing the plasma physics during an experiment. Therefore, they must be obtained as soon as possible. However, the execution of TSMAP requires computational power, and the performance of a typical personal computer is not high enough to keep up with the 3-min plasma discharge cycle. To increase the performance, the authors use a parallel computing approach. Because the fitting calculation for each time is independent, the calculations for different times can be executed simultaneously. Using this approach, the authors increased the performance by 25 times, achieving a 25-s execution time.