Analysis of the Efficiency of Using Computers with an Array-parallel Architecture for Processing Spectral Data of Astronomical Objects

The work is devoted to the methods of parallel processing of spectral data of astronomical objects, which is an important aspect in many scientific fields, in particular in astronomy when studying physical processes and chemical composition of space objects. The relevance and practical significance of the work is caused by the increase in the amount of spectral data obtained from modern radio telescopes, which requires new approaches to their processing. The use of GPU-based massively parallel computers provides fast and efficient processing of large data sets, which is critical for modern astronomical research. This paper proposes an improved model for processing spectral data of astronomical objects using parallel computing on GPUs. The aim of this paper is to analyse the efficiency of using computers with an array-parallel architecture for processing spectral data of astronomical objects. To achieve this goal, the following methods were analysed and used: morphological operations, methods of preliminary data normalisation, methods of load balancing during parallel processing of input data, methods of massively parallel computing, methods of creating windows and graphic elements. The study results showed that with large amounts of data, the GPU is more efficient than the CPU, providing more than 1.6 times the speedup. For example, when the initial dataset size specified in the experiment was increased, the processing time on the GPU (0.091471 seconds) was significantly less than on the CPU (0.126231 seconds). Further research will focus on improving the algorithms for parallel processing of spectral data, expanding the functionality of the tools used, and adapting the model for new applications in astronomy.