关键词: 多核环境, 并行计算, GNSS数据处理, 分块矩阵, 大规模GNSS网平差

Abstract: Multi-core processors have becoming the mainstream of current computer architecture, and multi-core parallel computing technology and its application has gotten increasing concerns. Generally, traditional GNSS data processing procedures are written for single processor architecture. This paper studies the parallel algorithm for computationally intensive tasks of the current GNSS data processing with multiple time periods or many stations under multi-core environment. The hot spot calculated tasks which GNSS data processing deal 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 GNSS data processing 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.

Key words: multi-core environment, parallel computing, GNSS data processing, modular matrix, large scale GNSS network adjustment