项目名称: 基于GPU的脉冲星宽带观测的相干消色散研究
项目编号: No.11303093
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 徐永华
作者单位: 中国科学院云南天文台
项目金额: 29万元
中文摘要: 本项目针对当前脉冲星宽带相干消色散接收机的研制中面临的海量数据的实时处理问题,开展脉冲星宽带的相干消色散研究工作。其相比于非相干消色散,可以提高计时精度、观测精度和灵敏度等。针对脉冲星的宽带观测带来海量数据实时处理需求,以及传统解决方案的性能和成本等带来巨大的挑战,本课题利用新兴的运算平台GPU实现脉冲星1GHz观测带宽,8比特量化数据的准实时相干消色散。采用CPU/GPU异步计算模式和大规模并行数据的通用数据结构,充分挖掘CPU/GPU的计算能力,实现CPU和GPU的真正的并行运算;研究GPU 中线程分配和调度,优化相干消色散算法,实现算法进一步加速;利用大规模并行数据的通用数据结构,解决GPU负载平衡度的问题。
中文关键词: GPU;相干消色散;异步计算模式;;
英文摘要: This project for the current development of broadband coherent pulsar de-dispersion receiver in the face of huge amounts of data real-time processing of problem,coherent de-dispersion algorithm research. Compared with the noncoherent de-dispersion,it can improve the timing precision, observation precision and sensitivity, etc. In view of the pulsar broadband observations make huge amounts of data real-time processing requirements, and the performance of the traditional solution and cost bring huge challenges, in this project we will design the real-time coherent de-dispersion algorithm of 1 GHZ bandwidth observation and 8 bits of quantitative data. Using CPU/GPU asynchronous computation pattern and generic data structure of large-scale parallel data, fully tap the CPU/GPU computing ability, realize the real CPU and GPU parallel computing; Studies GPU threads in the allocation and scheduling, optimal coherent de-dispersion algorithm, the algorithm further speed up; Using a generic data structure to effectively manage large scale parallel, solve the problem of the GPU load balance.
英文关键词: GPU;Coherent De-dispersion;Asynchronous Computing Pattern;;