High-resolution Solar Image Reconstruction Method Based on Multi-process Parallel Acceleration

At present, speckle interferometry and speckle masking are often used to reconstruct the mode and phase of the target in high-resolution solar image reconstruction. However, due to a large amount of grouped and partitioned data and the complexity of the algorithm, it is difficult to meet the requirements of real-time reconstruction. In order to alleviate the pressure of data processing, based on the existing CPU/GPU hybrid computing method of single block data, a method of allocating multi-component block data to GPU through multi-process and parallel processing is proposed. The experiment shows that the method based on a multi-process parallel acceleration method can improve the utilization efficiency of CPU and GPU resources, enable GPU to process multi-block data at the same time, and significantly improve the speed of image block processing with an acceleration ratio of about 4.7. Relevant studies can provide a reference for the parallel processing of astronomical data.