The software-defined fast post-processing for GEM soft x-ray diagnostics in the Tungsten Environment in Steady-state Tokamak thermal fusion reactor

Abstract

This article presents a novel software-defined server-based solutions that were introduced in the fast, real-time computation systems for soft X-ray diagnostics for the WEST (Tungsten Environment in Steady-state Tokamak) reactor in Cadarache, France. The objective of the research was to provide a fast processing of data at high throughput and with low latencies for investigating the interplay between the particle transport and magnetohydrodynamic activity. The long-term objective is to implement in the future a fast feedback signal in the reactor control mechanisms to sustain the fusion reaction. The implemented electronic measurement device is anticipated to be deployed in the WEST. A standalone software-defined computation engine was designed to handle data collected at high rates in the server back-end of the system. Signals are obtained from the front-end field-programmable gate array mezzanine cards that acquire and perform a selection from the gas electron multiplier detector. A fast, authorial library for plasma diagnostics was written in C++. It originated from reference offline MATLAB implementations. They were redesigned for runtime analysis during the experiment in the novel online modes of operation. The implementation allowed the benchmarking, evaluation, and optimization of plasma processing algorithms with the possibility to check the consistency with reference computations written in MATLAB. The back-end software and hardware architecture are presented with data evaluation mechanisms. The online modes of operation for the WEST are discussed. The results concerning the performance of the processing and the introduced functionality are presented.

Parallel computing in soft X‐rays plasma diagnostic systems for thermal fusion reactors—feasibility studies for GPUs

“…These positive results render the implementation of the proposed architecture possible. The suggestions to further investigate GPUs in the previous studies were correct.…”

“…The front‐end impact on the back‐end analysis has been discussed by Krawczyk et al The development of the heterogeneous architecture of FPGAs and HPC servers has been documented by Krawczyk et al with the most relevant advantages of such an approach. The real‐time modes of operation have been presented by Krawczyk et al…”

“…That impacts optimization, performance, evaluation, and the development of parallel solutions. The initial runtime implementation is presented in with the algorithm using OpenMP, as depicted in Figure …”

“…In the upper scheme, the group of CPUs is used. This approach has already been successfully utilized . In the bottom scheme of Figure , the CPUs only handle the control over transmission (marked as a double‐ended arrow).…”

FPGA and Embedded Systems Based Fast Data Acquisition and Processing for GEM Detectors

“…Further limitations are introduced by the ES part. Latest analyzes [37,38] have shown that a system based on Xeon e5 S2600 v4 CPU with 10 cores is able to process up to 6.9 million photons per second. However, the system is still under development, and further improvement both regarding the PCIe bandwidth and processing throughput is expected.…”