论文部分内容阅读
本文在实现重力全张量梯度数据重加权聚焦反演的基础上,利用MPI与CUDA实现了算法的并行运算,以及二者联合的混编并行程序,引入并行计算中的程序性能指标,对算法性能进行分析和综合对比,总结出一套评价规则用于并行算法程序的性能分析。此规则给出算法各自的性能评估方法,有助于客观全面地评价程序的性能,从程序开发和算法实现的角度帮助评估和提高算法的并行效率。同时我们将该套程序和评价规则用于理论模型和Vinton盐丘的实测梯度数据中,不仅反演出良好的密度结果,还获得了较高的运算效率,验证了所使用并行算法在重力全张量梯度数据反演中的可行性,根据评价规则认为程序均具有良好的运行效率。
In this paper, based on the gravity weighted total-tensor gradient reconstruction of the weighted data, based on the parallel algorithm using MPI and CUDA parallel computing, as well as the combination of the two parallel programs, the introduction of parallel computing program performance indicators, the algorithm Performance analysis and comprehensive comparison, summed up a set of evaluation rules for parallel algorithm performance analysis program. This rule gives the performance evaluation method of each algorithm, which helps to evaluate the performance of the program objectively and comprehensively, and helps to evaluate and improve the parallelism of the algorithm from the perspective of program development and algorithm implementation. At the same time, we apply this set of procedures and evaluation rules to the theoretical model and the measured gradient data of Vinton salt dome, not only the good density results are inversed but also the higher computational efficiency. According to the evaluation rules, the program is considered to have good operation efficiency.