-
摘要: 三维磁流体力学(MHD)数值模拟是用来研究日冕和太阳风最常用的方法之一, 其中将计算得到的日冕电子数密度转化为日冕偏振亮度(Polarization Brightness, PB)是与观测对比的重要方法. 由于待转换电子数据网格密度、PB数据网格密度和计算模型的复杂度, 使得日冕偏振亮度的计算比较耗时, 利用单CPU计算无法达到近实时转换日冕偏振亮度的要求, 从而影响了数值模拟的验证效率. 本文在CPU/GPU环境下, 利用CUDA编程技术, 提出了一个日冕偏振亮度并行计算模型. 实验结果表明, 该模型比CPU上的串行模型计算速度提高了31.86倍, 达到了近实时模拟与观测数据比对的计算要求.Abstract: Applied Research, Chinese Academy of Sciences, Beijing 100190)
The three-dimensional numerical Magnetohydrodynamics (MHD) simulation is one of the methods to study the corona and solar wind. Comparison of the Polarization Brightness (PB) in the low corona between simulation results and observation is important to validate the MHD models. Because of the massive data and the complexity of the PB model, the computation will take too much time on single CPU (or core), and can not visualize the PB in nearly real time, so that it affects the validation efficiency of numerical modeling. A new parallelized model based on CPU/GPU with Compute Unified Device Architecture (CUDA) to convert MHD simulation density to PB is presented in this paper. Aiming at the most time-consuming part of the serial PB model, it modifies the model from serial process to parallel process based on CUDA to improve its computation efficiency. It shows that the new model can speed up 31.86 times than the serial model based on CPU, and is able to compare the simulation with observation in nearly real time. -
-
计量
- 文章访问数: 2969
- HTML全文浏览量: 115
- PDF下载量: 958
-
被引次数:
0(来源:Crossref)
0(来源:其他)