The Research On GPU Parallizing Of AST3 Images Co-addition Module Swarp

AST3 is the abbreviation of Three Schmidt Telescopes that will be installed at Kunlun South Pole Station which is located at Antarctic Dome A. Images sbustraction metering is one of three common methods in Astronomy metering, an important step of the images substraction metering is producting reference image, which is co-added by several astronomy images. Swarp is a most common tool for astronomy images co-addition that is used in AST3 pipeline.This paper meticulous does a research on the Swarp pipeline on the base of the pipeline of images processing for AST3 and discusses the principle of Swarp modules. Acording to run-time bottleneck analyzing, the most time consuming functions are found. The pseudocode of these founctions is studied, on above these studies, this paper provides three plans to accelerate the co-addition, two with GPU and the others with openMP. After fufils these plans, the co-additon time and speed-up are analyzed with different input images. By the second GPU plan, 2/3 CPU serial run-time is cut off.To the convenience of the second time development, a makefile template with GPU paralleling is provided, with this template, the parallelized Swarp is easy to transplante to other astronomy image processing systems.