Table 1.
Performance of optimized kernel versions against the original code version. The deviation of the results present both the average and the maximum value measured.
Application | Wall time (s) |
Speed-up | Deviation |
||
---|---|---|---|---|---|
Orig. | Opti. | Mean | Max | ||
FDTD | 24.53 | 18.67 | 1.31 | 0.29% | 0.99% |
GOL | 170.47 | 93.58 | 1.82 | None | None |
Fluid sim. | 108.67 | 82.47 | 1.32 | 0.11% | 4.64% |
K-Means | 15.73 | 11.37 | 1.38 | 1.7% | 4.9% |