arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1607.01802astro-ph.IMastro-ph.SRphysics.comp-ph
收藏

混合自适应射线矩量法(HARM $^2$):一种基于自适应网格的辐射流体动力学高度并行方法

Hybrid Adaptive Ray-Moment Method (HARM$^2$): A Highly Parallel Method for Radiation Hydrodynamics on Adaptive Grids

Anna L. Rosen, Mark R. Krumholz, Jeffrey S. Oishi, Aaron T. Lee, Richard I. Klein

摘要:

我们提出了一种高度并行的多频混合辐射流体动力学算法,该算法结合了空间自适应的长特征方法的辐射场从点源与矩量法,处理由体积填充流体产生的漫辐射场。我们的混合自适应射线矩量法(HARM $^2$)基于补丁的自适应网格上运行,是兼容异步时间步进,并与任何时刻的方法。与以前的长特征方法相比,我们通过开发一种新的完全异步和非阻塞的通信算法,大大提高了自适应长特征方法的并行性能。作为这种改进的结果,我们的实现达到了近乎完美的扩展到分布式内存机器上的$\mathcal{O}(10^3)$处理器。我们提出了一系列的测试,以证明该方法的准确性和性能。

英文摘要:

We present a highly-parallel multi-frequency hybrid radiation hydrodynamics algorithm that combines a spatially-adaptive long characteristics method for the radiation field from point sources with a moment method that handles the diffuse radiation field produced by a volume-filling fluid. Our Hybrid Adaptive Ray-Moment Method (HARM$^2$) operates on patch-based adaptive grids, is compatible with asynchronous time stepping, and works with any moment method. In comparison to previous long characteristics methods, we have greatly improved the parallel performance of the adaptive long-characteristics method by developing a new completely asynchronous and non-blocking communication algorithm. As a result of this improvement, our implementation achieves near-perfect scaling up to $\mathcal{O}(10^3)$ processors on distributed memory machines. We present a series of tests to demonstrate the accuracy and performance of the method.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑