Optimization Research On CLEAN Algorithm In ARL

The Square Kilometre Array,the world’s largest synthetic aperture radio telescope,plays a vital role in astronomical radio observations.Algorithm Reference Library is a candidate algorithm library for SKA.CLEAN algorithm is a very important algorithm in imaging part.An improved CLEAN algorithm,the Multi-Scale CLEAN algorithm,is used in ARL.This algorithm can process exhibition source well,get a clearer image and reflect distribution of sky more realistically.In view of long running time of Multi-Scale CLEAN algorithm,in this paper,an MS-Clark CLEAN algorithm is proposed based on Multi-Scale CLEAN algorithm.Then the algorithm is implemented by GPU parallelization.Finally,M31 simulation data is used to test and analyze algorithm.Compared with the running time of Multi-Scale CLEAN algorithm in ARL without affecting image quality,the algorithm proposed in this paper has shorter running time and higher computational efficiency.Main research work of this paper is as follows:1、An MS-Clark CLEAN algorithm is proposed.Based on Multi-Scale CLEAN algorithm,algorithm combines the advantages of Clark CLEAN algorithm.Under premise of ensuring image quality,it is expected to reduce running time of algorithm and improve efficiency of operation.2、Research on GPU parallelization of CLEAN algorithm.In order to improve computational efficiency of algorithm,this paper uses the advantages of current GPU to design CUDA kernel function running on GPU device for the parallel part of MS-Clark CLEAN algorithm.And use PyCUDA to achieve parallelization of algorithm.By parallelizing algorithm and time-consuming analysis of function,parallel and time-consuming parts of algorithm can be found to improve the parallelization effect.Experiments show that the algorithm based on GPU parallelization can improve computational efficiency.