| 引用本文: | 张曦,孙旭,郭晓虎,等.面向GPU的非结构网格有限体积计算流体力学的图染色方法优化.[J].国防科技大学学报,2022,44(5):24-34.[点击复制] |
| ZHANG Xi,SUN Xu,GUO Xiaohu,et al.Optimizations of graph coloring method for unstructured finite volume computational fluid dynamics on GPU[J].Journal of National University of Defense Technology,2022,44(5):24-34[点击复制] |
|
| |
|
|
| 本文已被：浏览 5178次 下载 3687次 |
| 面向GPU的非结构网格有限体积计算流体力学的图染色方法优化 |
| 张曦1,孙旭1,郭晓虎2,杜云飞1,卢宇彤1,刘杨3 |
|
(1. 中山大学 计算机学院(软件学院), 广东 广州 510006;2. 哈璀国家超算中心 达斯伯里实验室, 英国 沃林顿 WA4 4AD;3. 中国空气动力研究与发展中心, 四川 绵阳 621000)
|
| 摘要: |
| 采用图染色方法解决通量累加和局部最大压力计算引起的两种典型资源竞争问题,并通过共享内存的使用、体编号和面编号的重排、面数据的重排三种策略优化图染色方法。针对应用在空气动力学多种规模的三维网格,分别采用双精度和单精度操作数,在Nvidia Tesla V100和K80 GPU上,展开性能测试。结果表明:共享内存的优化效果不明显；体编号和面编号重排降低了图染色方法的计算性能；面数据重排可以有效地优化图染色方法；计算性能在V100上提高20%左右,在K80上提高15%左右。 |
| 关键词: 非结构网格 有限体积 图形处理器 资源竞争 图染色 |
| DOI：10.11887/j.cn.202205003 |
| 投稿日期：2020-11-09 |
| 基金项目:国家重点研发计划资助项目(2016YFB0200902)；国家数值风洞工程资助项目(NNW2019ZT6-B18)；广东省引进创新创业团队资助项目(2016ZT06DZ11) |
|
| Optimizations of graph coloring method for unstructured finite volume computational fluid dynamics on GPU |
| ZHANG Xi1, SUN Xu1, GUO Xiaohu2, DU Yunfei1, LU Yutong1, LIU Yang3 |
|
(1. School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, China;2. STFC Daresbury Laboratory, Hartree Centre, Warrington WA4 ;4.AD, UK;3. China Aerodynamics Research and Development Center, Mianyang 621000; China)
|
| Abstract: |
| Graph coloring was used to address resource competition for the two typical computing procedures, including the flux summation and the calculation of local maximum pressure. There were three optimizations applied on graph coloring including shared memory, the reordering of volume and face indices, and the reordering of face variables. The 3D aerodynamics application with a series of mesh sizes was used in the performance test by double and single precision floating point operations on GPU Nvidia Tesla V100 and K80. The performance tests show that the shared memory is not obvious in performance. Furthermore, the reorder of volume and face indices reduces the performance of graph coloring.It is found that the reorder of face variables can increase performance remarkably. Specifically, the performance of graph coloring is increased by around 20% on V100 and 15% on K80. |
| Keywords: unstructured mesh finite volume graphic processing units race condition graph coloring |
|
|
|
|
|
