OpenMP并行计算在全球电离层反演中的应用

对全球电离层反演数据处理中的计算密集型任务进行分析，针对数据预处理、组建法方程矩阵、参数预消除和法方程矩阵求逆等主要模块设计了基于OpenMP（Open Multi-Processing）的并行计算方案。该实验方案在单台服务器下实施，通过算例验证了本文并行计算方案的有效性和可靠性。实验结果表明:采用并行计算后，全球电离层快速解执行时间只需要10~13 min，计算速度加快了约6倍；最终解执行时间只需要39~47 min，计算速度加快了约5倍。本文全球电离层模型精度约为2.8~3.8 TECU，最终解模型精度相比快速解精度提高了约0.2 TECU，与IGS各个分析中心电离层产品精度基本相当。

Application of Parallel Computing with OpenMP in Global Ionosphere Mapping

Computation-intensive tasks in the data processing for global ionosphere mapping are analyzed. A parallel algorithm with OpenMP technique is proposed for data preprocessing incorporating normal equation matrix stacking, parameter pre-elimination, and normal equation matrix inversion. A validation experiment was executed on a single server. Results indicate that the computing time for rapid global ionosphere mapping is only about 10~13 minutes with a speedup of about six, the time for final global ionosphere mapping is only about 39-47 minutes and the speedup is about five. The precision for global ionosphere products are about 2.8~3.8 TECUs compared with IGS final products, and the final solutions are at higher precision, about 0.2 TECUs in contrast to rapid solutions. The parallel algorithm proposed in this paper is very useful for the optimal testing of global ionosphere mapping, when reprocessing long time span data extending over periods such as 10 years, as well as for the real-time ionosphere mapping.