From: Multistep schemes for solving backward stochastic differential equations on GPU
K | N | M | \(|y_{0,0}-y_{0}^{0}|\) | \(|z_{0,0}-z_{0}^{0}|\) | \(t_{CPU}\) | \(t_{GPU}\) | speedup | Memory |
---|---|---|---|---|---|---|---|---|
1 | 8 | 46 | 1.32E−02 | 1.95E−03 | 0.14 | 0.01 | 18.60 | 0.30 |
1 | 16 | 64 | 6.46E−03 | 4.64E−03 | 0.53 | 0.01 | 45.96 | 0.30 |
1 | 32 | 92 | 3.18E−03 | 3.24E−03 | 2.12 | 0.04 | 47.55 | 0.30 |
1 | 64 | 128 | 1.57E−03 | 1.85E−03 | 8.73 | 0.17 | 50.33 | 0.30 |
2 | 8 | 78 | 5.90E−04 | 8.49E−03 | 0.62 | 0.01 | 44.08 | 0.30 |
2 | 16 | 128 | 8.31E−04 | 3.46E−03 | 3.56 | 0.07 | 49.93 | 0.31 |
2 | 32 | 216 | 5.84E−04 | 1.44E−03 | 21.96 | 0.42 | 51.75 | 0.34 |
2 | 64 | 364 | 3.39E−04 | 6.39E−04 | 131.09 | 2.57 | 50.95 | 0.42 |
3 | 8 | 128 | 3.94E−04 | 1.43E−03 | 2.12 | 0.04 | 49.40 | 0.32 |
3 | 16 | 256 | 6.75E−05 | 2.08E−04 | 21.05 | 0.41 | 51.83 | 0.38 |
3 | 32 | 512 | 9.72E−06 | 2.79E−05 | 187.79 | 3.61 | 52.04 | 0.65 |
3 | 64 | 1024 | 1.30E−06 | 3.60E−06 | 1719.55 | 30.54 | 56.30 | 1.66 |
4 | 8 | 128 | 1.90E−04 | 8.09E−04 | 2.41 | 0.05 | 45.19 | 0.33 |
4 | 16 | 256 | 3.78E−05 | 1.24E−04 | 25.99 | 0.51 | 51.18 | 0.40 |
4 | 32 | 512 | 5.69E−06 | 1.67E−05 | 239.73 | 4.63 | 51.80 | 0.75 |
4 | 64 | 1024 | 7.76E−07 | 2.17E−06 | 2279.06 | 40.10 | 56.83 | 2.09 |
5 | 8 | 128 | 1.39E−04 | 7.49E−04 | 2.39 | 0.05 | 45.06 | 0.35 |
5 | 16 | 256 | 3.65E−05 | 1.30E−04 | 30.16 | 0.59 | 51.27 | 0.43 |
5 | 32 | 512 | 6.00E−06 | 1.83E−05 | 292.10 | 5.67 | 51.54 | 0.86 |
5 | 64 | 1024 | 8.48E−07 | 2.42E−06 | 3049.33 | 48.75 | 62.55 | 2.52 |
6 | 8 | 128 | 8.13E−05 | 5.97E−04 | 2.21 | 0.04 | 50.61 | 0.34 |
6 | 16 | 256 | 3.07E−05 | 1.18E−04 | 32.93 | 0.64 | 51.76 | 0.46 |
6 | 32 | 512 | 5.49E−06 | 1.73E−05 | 341.15 | 6.46 | 52.81 | 0.97 |
6 | 64 | 1024 | 8.00E−07 | 2.32E−06 | 3394.13 | 57.56 | 58.96 | 2.94 |