Research Article
An Optimized Parallel FDTD Topology for Challenging Electromagnetic Simulations on Supercomputers
Table 2
Comparisons of virtual topology, amount of communication, and computation time on NSCC-TJ.
| CPU cores | Virtual topology () | Amount of communication | Computation time (in seconds) |
| 12 | 3 × 2 × 2 | 2520000 | 6466.68 |
| 16 | 4 × 2 × 2 | 2880000 | 6140.56 |
| 32 | 4 × 4 × 2 | 3600000 | 2858.97 |
| 64 | 8 × 8 × 1 | 5040000 | 1376.85 | 8 × 4 × 2 | 5040000 | 1379.92 | 16 × 2 × 2 | 7200000 | 1824.37 |
| 96 | 8 × 6 × 2 | 5760000 | 946.52 | 8 × 4 × 3 | 6480000 | 1500.54 | 12 × 4 × 2 | 6480000 | 1551.95 | 16 × 3 × 2 | 7560000 | 1679.42 |
| 120 | 6 × 10 × 2 | 6480000 | 702.16 | 10 × 6 × 2 | 6480000 | 808.65 | 5 × 12 × 2 | 6840000 | 713.67 | 12 × 5 × 2 | 6840000 | 863.65 | 5 × 6 × 4 | 7560000 | 724.90 | 6 × 5 × 4 | 7560000 | 1076.41 | 15 × 4 × 2 | 7560000 | 1096.52 |
|
|