浅水方程的并行化求解  

Parallel Simulation of Shallow Water Flow

作　　者：周洋[1] 张景新[1] 

机构地区：[1]上海交通大学船舶海洋与建筑工程学院,上海200240

出　　处：《力学季刊》2013年第4期607-613,共7页Chinese Quarterly of Mechanics

基　　金：上海市重点学科建设项目(B206)

摘　　要：随着计算机硬件的快速发展,空间高分辨率的数值模拟得以在河口海岸带的大尺度水动力学研究中获得应用.结合当今共享内存多处理器系统SMP(Shared-memory MultiProcessor)计算机的发展趋势,针对模拟潮汐流动的浅水方程采用OpenMP进行了并行化.研究了不同线程数下计算规模、并行粒度以及线程绑定技术对于并行效率的影响,为特定规模的计算选择最优线程数提供了依据.Spatial high resolution simulation can be used to study the large-scale hydrodynamics in estuarine and near-shore regions as the rapid development of high performance computer. Seeing a rising trend in the development of SMP structure computer, an in-house code simulating tide flow by solving shallow water equation was parallelized with OpenMP. The impact factors on parallel performance, such as computational scale, parallel granularity and binding thread in several cases with different numbers of threads were studied, the results reveal a feasible way to determine a choice of optimal thread number referring to a certain computational scale.

关 键 词：SMP OPENMP CG 浅水方程 并行计算 

分 类 号：O352[理学—流体力学]