项目名称: 多GPU并行的热/化学反应非平衡N-S方程求解算法研究
项目编号: No.11302065
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 张兵
作者单位: 合肥工业大学
项目金额: 26万元
中文摘要: 热/化学反应非平衡N-S方程的数值求解问题,具备运算复杂度高、浮点运算密集等特点,采用浮点性能极高的图形处理器(GPU)可以大大提高分析速度。本项目以多GPU并行的热/化学反应非平衡N-S方程求解算法为研究对象,针对三个关键问题:GPU上的复杂算法的寄存器优化方法、GPU上非结构网格的高效数据存储结构和访问方式优化、适用于多GPU的高效并行算法,采用编译优化、最大化存储器合并访问、染色通信等技术开展研究。最终实现多GPU上大规模热/化学反应非平衡N-S方程求解问题的快速计算,为高超声速流动的高效高精度数值模拟提供重要手段。
中文关键词: 图形处理器;并行计算;热化学反应非平衡;计算流体力学;高超声速流动
英文摘要: High computational complexity and intensive floating point arithmetic are the two major features of numerical simulation of thermo-chemical nonequilibrium Navier-Stokes equations. The solving process can be speed up enormously by using modern Graphic Process Units (GPU) which has high floating point performance. This project is focused on the parallel computational methods for solving thermo-chemical nonequilibrium Navier-Stokes equations using multi-GPU. The register optimization method of complex algorithm, the efficient data structure and memory access optimization of unstructured mesh, the parallel algorithm for multi-GPU will be studied in deep by using compiler optimization, maximize memory throughput, and coloration packet communication techniques. A rapid simulation will be achieved for solving large-scaled thermo-chemical nonequilibrium Navier-Stokes equations problem using multi-GPU, and an efficient and high accurate numerical method be provided for hypersonic flow simulation.
英文关键词: Graphics Processing Unit;Parallel Computing;Thermo-Chemical Nonequilibrium;Computational Fluid Dynamics;Hypersonic Flow