Table 1.
Computing step | Linear solver | Non-linear solver |
---|---|---|
GPU (CPU, speedup) | GPU (CPU, speedup) | |
Boltzmann | 9.060 s (1 min 31 s, ×10.05) | 8.83 s (1 min 28 s, ×9.96) |
Iteration | 0.015 s (0.18 s, ×10.60) | 0.035 s (0.44 s, ×12.57) |
Total | 10.250 s (1 min 38 s, ×9.61) | 10.14 s (1 min 37 s, ×9.55) |
Note: The execution time is reported per computing step: ‘Boltzmann’ includes the overall time spent for Laplace and Boltzmann updates. ‘Iteration’ is the time spent in a single iterative step. ‘Total’ includes all the iterations and the initialization of the GPU card.