EDLD-Tool: A real-time GPU-based tool to stream and analyze energy-dispersive Laue diffraction of BIG Data sets collected by a pnCCD

摘要

Energy-dispersive Laue diffraction (EDLD) is a tool for the characterization of singlecrystalline and polycrystalline materials. Using a two-dimensional energy-dispersive detector, both the angular positions and the diffracting energies of the Laue spots can be analysed without additional information, allowing for fast indexing, determination of the crystal structure and the respective lattice parameters. Running the detector at ≈ 400 Hz, a typical data set (≈ 10 min) taken by the single photon counting mode has a size of ten Gigabytes. Up to now, problems such as data transfer, data storage, data reduction and on time data analysis are drawbacks for wider application of this technique. A fast and effective algorithm for processing of these BIG Data is required to overcome the drawbacks and to allow for quality assessment of the running experiment in real time. This paper presents a GPU based tool for energy-dispersive Laue diffraction (EDLD) experiments, named "EDLD-Tool", providing the optimization of geometric parameters of the experiment, autoindexation, online steering, error detection and determination of all crystal parameters considering pnCCD data taken from a single crystal. This tool is exploiting parallel computing technology of the GPU and makes use of many scientific, high performance libraries (i.e. OpenCV, Root-Cern, Eigen and others). As a result, the developed tool allows for data processing in the time frame of few seconds compared to the previous analysis system which requires few hours to process the same amount of data.