多物理并行数值模拟中的两层紧耦合联接算法

Two-Level Tightly Concatenating Algorithm for Multiphysics Parallel Numerical Simulations

复杂物理现象通常由多类复杂的物理过程紧耦合构成 ,其数值模拟也通常由适用不同物理过程的多类并行应用程序紧耦合完成 .如何设计这些物理过程之间的联接算法 ,既要保证程序之间数据传递的高效 ,又要保证程序各自运行和总体模拟的高效 ,还要保证程序各自开发的独立 ,是一个值得研究的课题 .该文基于广泛应用于高温高压多物理研究中的辐射流体力学和中子输运多物理并行数值模拟 ,在非结构网格上 ,提出了两种联接算法 :完全松散联接算法和两层紧耦合联接算法 ,前者侧重于实现程序各自运行的高效和开发的独立 ,后者在前者的基础上 ,还权衡了数据传递和总体模拟的高效 .在两台并行机的数百个处理机上 ,通信复杂度分析和数值实验结果表明两个算法均是有效的 ,可推广适用于辐射或中子输运与其他流体力学的多物理并行数值模拟应用中 .特别地 ,两层紧耦合联接算法是高效可扩展的 ,取得了近似最优的并行性能 .

Multiphysics parallel numerical simulations are usually essential to simplify researches on complex physical phenomena in which several physics processes are tightly coupled. It is very important on how to concatenate those coupled physics processes for fully scalable parallel simulation. Meanwhile, three objectives should be balanced, the first is efficient data transfer among simulations, the second and the third are efficient parallel executions and simultaneously developments of those simulation codes. This paper presents two concatenating algorithms for multiphysics parallel numerical simulations coupling radiation hydrodynamics with neutron transport on unstructured grid. The first algorithm, Fully Loosely Concatenation (FLC), focuses on the independence of code development and the independence running with optimal performance of code. The second algorithm, Two Level Tightly Concatenation (TLTC), focuses on the optimal tradeoff among above three objectives. Theoretical analyses for communicational complexity and parallel numerical experiments on hundreds of processors on two parallel machines have shown that these two algorithms are efficient and can be generalized to other multiphysics parallel numerical simulations coupling radiation or neutron transport with hydrodynamics. Especially, algorithm TLTC is linearly scalable and has achieved the optimal parallel performance.