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 | 2.71E−03 | 1.21E−02 | 0.15 | 0.00 | 31.96 | 0.33 |
1 | 16 | 64 | 1.16E−03 | 6.22E−03 | 0.61 | 0.01 | 49.03 | 0.34 |
1 | 32 | 92 | 5.62E−04 | 3.18E−03 | 2.49 | 0.04 | 56.82 | 0.35 |
1 | 64 | 128 | 2.67E−04 | 1.61E−03 | 10.29 | 0.17 | 58.93 | 0.34 |
2 | 8 | 78 | 9.09E−05 | 2.31E−03 | 0.65 | 0.01 | 44.33 | 0.35 |
2 | 16 | 128 | 5.06E−05 | 1.34E−03 | 4.08 | 0.07 | 57.72 | 0.35 |
2 | 32 | 216 | 1.75E−05 | 7.22E−04 | 26.64 | 0.42 | 63.82 | 0.38 |
2 | 64 | 364 | 8.00E−06 | 3.73E−04 | 153.21 | 2.51 | 60.96 | 0.46 |
3 | 8 | 128 | 6.79E−06 | 4.23E−06 | 2.19 | 0.04 | 58.92 | 0.36 |
3 | 16 | 256 | 3.73E−08 | 2.67E−07 | 22.90 | 0.38 | 61.02 | 0.42 |
3 | 32 | 512 | 7.56E−08 | 6.22E−08 | 214.17 | 3.41 | 62.74 | 0.69 |
3 | 64 | 1024 | 1.69E−08 | 9.01E−09 | 1911.74 | 29.26 | 65.34 | 1.70 |
4 | 8 | 128 | 1.07E−07 | 2.40E−06 | 2.29 | 0.04 | 57.21 | 0.36 |
4 | 16 | 256 | 5.16E−07 | 1.22E−07 | 28.10 | 0.46 | 61.37 | 0.45 |
4 | 32 | 512 | 4.19E−08 | 5.90E−08 | 275.75 | 4.37 | 63.16 | 0.79 |
4 | 64 | 1024 | 1.11E−08 | 6.46E−09 | 2509.67 | 38.55 | 65.10 | 2.13 |
5 | 8 | 128 | 6.50E−06 | 1.06E−06 | 2.17 | 0.04 | 57.84 | 0.37 |
5 | 16 | 256 | 1.53E−04 | 1.07E−07 | 32.14 | 0.53 | 60.92 | 0.47 |
5 | 32 | 512 | 2.84E−08 | 6.07E−08 | 333.62 | 5.31 | 62.78 | 0.90 |
5 | 64 | 1024 | 1.04E−08 | 5.54E−09 | 3083.75 | 47.94 | 64.32 | 2.56 |
6 | 8 | 128 | 6.70E−05 | 9.72E−07 | 1.77 | 0.03 | 57.63 | 0.38 |
6 | 16 | 256 | 4.27E−07 | 7.13E−08 | 35.05 | 0.58 | 60.59 | 0.50 |
6 | 32 | 512 | 7.56E−07 | 7.14E−08 | 387.05 | 6.19 | 62.56 | 1.01 |
6 | 64 | 1024 | 1.02E−08 | 4.73E−09 | 3666.74 | 57.06 | 64.26 | 2.99 |