Table 1 Speed-up (ratio of execution times) gained by the CUDA implementation with respect to the previous Python version. Benchmark on platform 1
Speed-up ± SD | total | kernel |  |
|---|---|---|---|
Combination length | 3 | 2 | 3 |
GN | 18.7 ±0.6 | 19.3 ± 1.8 | 18.3 ± 0.6 |
BK | 104.7 ±4.2 | 12887.5 ±1846.3 | 21993.4 ±1479.2 |
NF | 76.5 ±2.2 | 66195.2 ±1035.1 | 11877.1 ± 433.5 |
NFG | 84.0 ±3.3 | 6665.7 ±1105.0 | 13164.5 ± 595.6 |