射电脉冲星搜索的优化方法

Parallel optimization of the pulsar search pipeline

天文学计算通常具有数据量大、计算量多的特点.平方公里阵列射电望远镜(Square Kilometre Array,SKA)等装置的建设和高性能计算平台的发展促进了天文学与高性能计算之间的联系.脉冲星搜索是SKA的主要科学方向之一.本文介绍了一种基于OpenMP多线程和Multiprocessing多进程技术加速脉冲星搜索管线的方案,提出了一种解决负载不平衡问题的方法,并成功地将优化后的管线安装于中国SKA区域中心原型机的x86和ARM计算节点.在默奇森大视场阵列望远镜(Murchison Widefield Array,MWA)的3个脉冲星搜索样例上的测试显示,与原始单核心管线相比,优化后的管线在x86和ARM节点上分别获得10.4–12.2倍和24.5–25.8倍的加速,并行效率达到37.1%–43.6%和24.5%–26.9%.并且ARM节点比x86节点的计算快1.1–1.3倍,显示出国产计算设备在SKA数据处理方面的巨大潜力.这项应用优化成果,近期将用于加速MWA南天快速两米巡天(Southern-sky MWA Rapid Two-metre,SMART)项目的脉冲星搜寻工作.

The connection between astronomy and high-performance computing is becoming stronger with the development of innovative observing facilities such as the Square Kilometre Array(SKA)and the proposed innovative platform for big data and high-performance computing.Astronomical computation is characterized by a large data volume and massive parallelism,particularly for pulsar search,a leading scientific direction of the SKA.In this study,we present an approach for accelerating the pulsar search pipeline based on Open MP and multiprocessing techniques.We propose a method for solving the load imbalance problem and have successfully installed the pipeline on x86 and ARM compute nodes on the China SKA Regional Centre prototype(CSRC-P).The performance evaluation from tests on the Murchison Widefield Array(MWA)VCS observations shows that our optimization method works well on the x86 and ARM nodes,improving the relative speedup by factors of 10.4–12.2 and 24.5–25.8,respectively,compared with the original single-thread approach.The ARM platform was 1.1–1.3 times faster than the x86 platform in the tested cases,showing its great potential for SKA data processing.This optimized pulsar search pipeline deployed on the CSRC-P will be used for the pulsar survey of the southern-sky MWA rapid two-meter program for various scientific goals,including pulsar timing arrays for gravitational wave detection.