摘要
对当前多核环境下多时段或者多测站的GNSS网平差数据处理所涉及的计算密集型任务并行算法进行研究。分析GNSS网平差数据处理涉及的热点计算任务,提出基于分块理论的矩阵乘法运算、矩阵分解运算等数值计算并行方法,对比了单核和多核环境下的计算时间。通过多个算例验证多核并行设计方法的有效性,利用.NET4.0框架下的Parallel Extensions实现相关并行设计。试验结果表明,大型GNSS网平差数据处理的多核并行计算能充分发挥多核体系带来的性能优势,极大提高资源利用效率和GNSS数据处理效率。
The study is concentrated on the parallel algorithm for computationally intensive tasks of the current data processing for GNSS network adjustment with multiple time periods or many stations under multi-core environment. The hot spot calculated tasks which data processing for GNSS network adjustment deals with are analyzed and numerical calculation parallel method for matrix multiplication and matrix decomposition based on modular matrix theory are proposed, Subsequently, the computational efficiencies of single-core and multi-core environment are compared. The Parallel Extensions under NET4. 0 framework is adopted and the validity of the multi-core concurrent design is demonstrated by some examples. The experimental results show that the multi-core parallel computing of large scale data processing for GNSS network adjustment can give full play to the advantage of the multi-core system performance and greatly improve the efficiency of GNSS data processing and the resource utilization,
出处
《测绘学报》
EI
CSCD
北大核心
2013年第5期661-667,共7页
Acta Geodaetica et Cartographica Sinica
基金
国家自然科学基金(41274015)
国家863计划(2013AA122501)
关键词
多核环境
并行计算
GNSS数据处理
分块矩阵
大型GNSS网平差
multi-core environment
parallel computing
GNSS data processing~ modular matrix~ large scale GNSSnetwork adjustment
