Research On High-resolution Solar Image Reconstruction Based On Collaborative Parallel Computing

With the continuous advancement of large-scale scientific projects,large-caliber,large-field astronomical observation instruments have been put into use one after another,the scale of data has become larger and larger,and the demand for complex solutions continues to rise.When the high-resolution solar image is reconstructed,it is difficult to efficiently meet the needs of real-time reconstruction by using SAA and Speckle masking.At present,the existing processing method still stays in the CPU single processor thread mode,which can easily lead to low utilization of the CPU and GPU,resulting in a waste of system resources.The key to solving this problem is to improve the existing algorithms and improve the collaborative parallel computing capabilities of the CPU and GPU.This thesis focuses on the most time-consuming links(image preprocessing,image alignment,image block processing)and different computing environments in Level1(SAA)and Level1+(speckle masking)and different computing environments in the SPCPC single process based on collaborative parallel optimization research,using collaborative parallel computing and multi-process(multi-threading)technology,focusing on the image block processing of the speckle mask method,an efficient collaborative parallel computing task model is established to optimize the timeconsuming steps of reconstruction Processing flow,research and propose SP-CPC,MPCPC collaborative parallel optimization,and acceleration methods.Besides,in order to make better use of multi-core CPUs,this paper also proposes MPI and MPI-CPC parallel acceleration methods.In this paper,the Ha channel data(10×100 frames,1024×024 pixels per frame)and Ti O channel data(10×50 frames,2160×2560 pixels)are compared with the existing SPCPC method.The MP-CPC parallel acceleration method is used in the image The two channels of initial alignment and image block processing can obtain speedups of 21.18 and 22.78,4.69 and 4.71 respectively.Also,the MPI and MPI-CPC acceleration methods have been experimentally verified,and good results have been obtained.Experiments show that MP-CPC makes full use of efficient collaborative parallel computing methods,multi-process technology,and the respective characteristics of GPU and CPU,effectively improving the utilization of CPU/GPU,and improving the speed of the reconstruction process.Its research can be parallel for astronomical data.The chemical treatment provides a reference for reference.