面向国产加速器的CFD核心算法并行优化

曹义魁,陆忠华,张鉴,刘夏真,袁武,梁姗

Parallel Optimization of CFD Core Algorithms Based on Domestic Processor

CAO Yikui,LU Zhonghua,ZHANG Jian,LIU Xiazhen,YUAN Wu,LIANG Shan

表1 ADI_kernel1_i和ADI_kernel5_i在不同线程块大小下的运行时间

Table 1 The running time of ADI_kernel1_i and ADI_kernel5_i under different thread block sizes

Block大小 ADI_kernel1_i ADI_kernel5_i 32*4*1 32*8*1 32*4*2 32*4*4 32*8*2 32*8*4 2.86 s 2.26 s 2.31 s 2.64 s 2.16 s 2.26 s 1.84 s 1.58 s 1.69 s 1.18 s 1.25 s 1.04 s