|
本帖最后由 stishovite 于 2025-7-17 16:21 编辑
参考GROMACS (2025.2 GPU版) 并行效率测试结果 - 分子模拟 (Molecular Modeling) - 计算化学公社帖子的内容,
测试了GROMACS 2025.2在单路纯CPU(AMD EPYC 9B14 96物理核心)上的并行效率,
结果显示,thread-MPI版效率高于OpenMPI版,效率最高的组合为 -ntmpi 48 -ntomp 2,可达157.21 ns/day,
但仍然远低于5080 显卡加速的 thread-MPI 版Gromacs,(-ntmpi 1 -ntomp 14,可达539.78ns/day)
--------------------------------------------------------------------------------------------------------------------------
Test Configuration | Perf (ns/day) | GPU Wait (%) | Bottleneck
--------------------------------------------------------------------------------------------------------------------------
screen-gmx_threadmpi-mode_cpu-np48-omp2 | 157.21 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np24-omp4 | 130.90 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np12-omp8 | 125.24 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np36-omp2 | 119.26 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np18-omp4 | 109.99 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np9-omp8 | 99.35 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np32-omp3 | 95.76 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp72 | 95.17 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np24-omp2 | 93.88 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp64 | 92.77 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp48 | 91.42 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np16-omp6 | 91.35 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np12-omp4 | 90.63 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp96 | 90.60 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np24-omp3 | 84.54 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np6-omp8 | 83.18 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp32 | 79.40 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np12-omp6 | 75.83 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np16-omp3 | 72.15 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np6-omp16 | 71.18 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np8-omp12 | 69.79 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp24 | 67.09 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np6-omp12 | 64.80 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np8-omp9 | 61.26 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np8-omp6 | 61.02 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np4-omp24 | 59.87 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np48-omp2 | 58.84 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np3-omp32 | 57.56 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np3-omp16 | 57.11 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp16 | 56.79 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np3-omp24 | 54.83 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np4-omp18 | 54.42 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np4-omp12 | 54.38 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np2-omp24 | 50.26 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np2-omp36 | 48.52 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np2-omp48 | 47.86 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp8 | 46.14 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np36-omp2 | 43.38 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp12 | 43.14 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np24-omp4 | 39.71 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np32-omp3 | 38.55 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np12-omp8 | 37.64 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np18-omp4 | 36.61 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np16-omp6 | 34.47 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np24-omp2 | 33.72 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np9-omp8 | 32.31 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np16-omp3 | 30.50 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np24-omp3 | 29.14 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np8-omp12 | 28.20 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np6-omp8 | 28.01 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np12-omp6 | 27.04 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np12-omp4 | 26.63 | N/A | CPU Only Run (N/A)
screen-gmx_threadmpi-mode_cpu-np1-omp4 | 25.83 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp32 | 25.80 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np6-omp16 | 24.71 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp24 | 24.59 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp48 | 24.11 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np6-omp12 | 23.89 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np8-omp9 | 23.82 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np8-omp6 | 23.43 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp16 | 22.00 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np4-omp18 | 21.84 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np4-omp12 | 21.46 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np3-omp16 | 21.13 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp64 | 20.95 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np4-omp24 | 20.92 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np3-omp24 | 20.22 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp72 | 19.50 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp12 | 18.35 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np2-omp24 | 18.02 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np2-omp36 | 17.18 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp8 | 17.08 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np3-omp32 | 14.85 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp4 | 13.04 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np2-omp48 | 12.45 | N/A | CPU Only Run (N/A)
screen-gmx_openmpi-mode_cpu-np1-omp96 | 12.37 | N/A | CPU Only Run (N/A)
--------------------------------------------------------------------------------------------------------------------------
|
|