TABLE 3.
Fastest block configurations to compute 20 000 time steps
Mesh size | GPU | CSW | towerDS | ||
---|---|---|---|---|---|
time | block size | time | block size | ||
256 | 1080Ti | 25.0849 | 64 × 4 × 1 | 20.0084 | 64 × 4 × 1 |
Titan X | 33.7281 | 64 × 4 × 1 | 34.8525 | 64 × 4 × 1 | |
P100 | 23.6460 | 64 × 4 × 1 | 14.0196 | 32 × 8 × 1 | |
512 | 1080Ti | 174.3350 | 64 × 4 × 1 | 165.5145 | 64 × 4 × 1 |
Titan X | 255.8197 | 64 × 4 × 1 | 238.4309 | 64 × 4 × 1 | |
P100 | 136.3548 | 128 × 4 × 1 | 112.2918 | 32 × 8 × 1 |