基于多进程并行加速的太阳高分辨图像重建方法  

High-resolution Solar Image Reconstruction Method Basedon Multi-process Parallel Acceleration

作　　者：邓涛 陈东[2] 代红兵[1] 王新华 Deng Tao;Chen Dong;Dai Hongbing;Wang Xinhua(School of Information Science and Engineering,Yunnan University,Kunming 650504,China;Yunnan Observatories,Chinese Academy of Sciences,Kunming 650216,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]云南大学信息学院,云南昆明650504 [2]中国科学院云南天文台,云南昆明650216 [3]中国科学院大学,北京100049

出　　处：《天文研究与技术》2021年第4期516-522,共7页Astronomical Research & Technology

基　　金：国家自然科学基金(U1831210)资助.

摘　　要：目前,太阳高分辨图像重建往往采用斑点干涉术和斑点掩模法重建目标的模和相位,由于分组分块数据量大,算法复杂等因素,难以满足实时重建的需求。为了缓解数据处理的压力,在现有的单组分块数据中央处理器/图形处理器混合计算方法的基础上,提出通过多进程将多组分块数据分配到图形处理器上同时并行处理的方法。实验结果表明,基于多进程并行加速方法可提高中央处理器和图形处理器的资源利用率,图形处理器能同时处理多组分块数据,显著提高图像分块处理的速度,加速比达到4.7左右。相关研究可以为天文数据并行化处理提供借鉴参考。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.

关 键 词：多进程 并行计算 图像重建 斑点掩模法 斑点干涉法 

分 类 号：P182.21[天文地球—天文学] TP751.2[自动化与计算机技术—检测技术与自动化装置]