Graphics processing unit-assisted density profile calculations in the KSTAR reflectometer.

Area of Science:

• Plasma physics
• Signal processing
• Computational physics

Background:

• Wavelet Transform (WT) is crucial for signal processing, particularly in plasma diagnostics.
• KSTAR's frequency modulation reflectometer uses WT for phase information extraction.
• Real-time density profile calculation is hindered by the time-intensive nature of WT.

Purpose of the Study:

• To detail the implementation of GPU-accelerated WT for the KSTAR reflectometer.
• To evaluate the performance improvements achieved through GPU utilization.
• To discuss the feasibility of real-time density profile calculation.

Main Methods:

• Implementing Wavelet Transform (WT) algorithms on Graphics Processing Units (GPUs).
• Optimizing data input/output (IO) processes to mitigate CPU-peripheral bottlenecks.
• Integrating GPU-accelerated WT into the KSTAR reflectometer data analysis pipeline.

Main Results:

• Significant reduction in data analysis time for WT processing.
• Demonstrated capability for real-time density profile calculation.
• Identified and addressed IO bottlenecks in the KSTAR data processing system.

Conclusions:

• GPU acceleration is effective in reducing WT computation time for reflectometer data.
• Real-time density profile calculation is achievable with optimized GPU implementation.
• Further optimization of data IO is crucial for fully realizing real-time processing capabilities.