This record is being processed for inclusion into GeoRef. It may not yet have been indexed, given a translated title, or checked by a GeoRef editor.
Analysis of efficiency of applying MKL and OpenMP polynuclear parallel algorithms to recovery of high degree Earth's gravity field
|Authors:||Chen Qiujie; Shen Yunzhong; Zhang Xingfu|
|Author Affiliations:||Primary: |
Tongji University, Department of Surveying and Geoinformatics, Shanghai, China
Guangdong University of Technology, China
|Volume Title:||Dadi Celiang yu Diqiu Dongli Xue Journal of Geodesy and Geodynamics|
|Source:||Dadi Celiang yu Diqiu Dongli Xue = Journal of Geodesy and Geodynamics, 32(5), p.118-123. Publisher: Gai Kan Bianjibu, Wuhan, China. ISSN: 1671-5942|
|Note:||In Chinese with English summary. 6 refs.; illus., incl. 5 tables|
|Summary:||The ordinary matrix operation algorithms are difficult to meet the needs of high degree gravity field recovery. In order to improve computational efficiency, this article applied OpenMP parallel computing libraries and MKL scientific computing libraries to high degree earth's gravity field recovery, which significantly improved the computational efficiency. Simulation results are as follows. 1)When making up of sub normal equation coefficient matrix with a single epoch as the solver unit, OpenMP algorithm can store normal equation coefficient matrix under the compressed triangle and OpenMP algorithm's overhead of memory and time are relatively acceptable. 2) When the author makes up of the high dimension's coefficient matrix with more than one epoch, and then forms the normal equation, MKL algorithm can make an efficient performance. 3) The efficiency of MKL inversion algorithm is much higher than OpenMP algorithm. 4) Utilization of the advantages of OpenMP and MKL algorithms can significantly improve the efficiency of high degree earth's gravity field recovery.|
|Subjects:||Algorithms; Data processing; Earth; Geodesy; Gravity field; Numerical analysis; Parallel processing|
|Copyright Information:||GeoRef, Copyright 2021 American Geosciences Institute.|
No Tags, Be the first to tag this record!