GPU并行计算技术在赫歇尔天文台远红外巡天数据处理中的应用

发布时间：2018-06-19 14:39

本文选题：GPU + 并行计算　；参考：《天文学进展》2013年04期

【摘要】：赫歇尔远红外银道面巡天数据由于其高分辨率和大尺度特性而具有海量性,使用图像处理单元(GPU)提高数据挖掘处理效率是海量数据处理中的一个重要环节。以GPU在实际研究工作中的使用方法和思路为着眼点,结合赫歇尔远红外数据处理过程中GPU的具体应用,介绍了GPU并行数据处理技术在星际介质物理性质和三维结构研究中的应用。对赫歇尔远红外数据的处理挖掘使用GPU加速的有两个部分:银道面温度图的每像素拟合和计算星际介质三维模型中单元格对星光的吸收。在这两个应用中,分别使用了CUDA和PyOpenCL调用GPU进行并行计算。这种使用GPU提高数据处理效率的思路与方法对与两个实例相类似的数据处理问题具有借鉴意义。
[Abstract]:Herschel far-infrared silver surface survey data is magnanimous because of its high resolution and large scale characteristics. Using image processing unit (GPU) to improve the efficiency of data mining is an important link in mass data processing. The application of GPU in the process of far infrared data processing is focused on the application of GPU in the practical research, and the specific application of GPU in the process of Herschel far infrared data processing is combined. The application of GPU parallel data processing technology in the study of physical properties and three-dimensional structure of interstellar media is introduced. The processing and mining of Herschel far infrared data are accelerated by GPU. There are two parts: fitting the temperature map of silver channel surface per pixel and calculating the absorption of starlight by cells in the three-dimensional model of interstellar medium. In these two applications, CUDA and PyOpenCL are used to call GPU for parallel computation. The idea and method of using GPU to improve the efficiency of data processing can be used for reference to the data processing problems similar to those of two examples.
【作者单位】：中国科学院国家天文台;中国科学院大学;
【基金】：中国科学院知识创新工程重要方向项目(KJCX2-YW-T20)
【分类号】：P111

【相似文献】