The new detour optimization is more inclined to rip up in order to optimize. The following is a selection of boards comparing the differences.
In all cases, the optimization improved the wiring, and sometimes resulted in a higher net completion rate. In most cases, the runtime increased, however in others, such as the Meggy Jr, the runtime decreased as the router closed in on a solution faster. The laminator had fewer successful nets, but those that were successful had better detours.
In some cases, applying the detour optimizations at an earlier stage produced even better results, as shown in the Meggy Jr and LED results. In some other boards the results were worse. The further possible optimization will be explored later with probabilistic algorithms and machine learning.
| Total Nets | Completed Nets | Wiring Length (inches) | Runtime (seconds) | |
|---|---|---|---|---|
| Autorouter | 123 | 123 | 60.7064 | << 1 |
| Toporouter (before optimization) | 123 | 123 | 52.7943 | 2.724 |
| Toporouter (detour optimization) | 123 | 123 | 49.3204 | 3.248 |
| Toporouter (best optimization) | 123 | 123 | 47.9989 | 3.542 |
| Total Nets | Completed Nets | Wiring Length (inches) | Runtime (seconds) | |
|---|---|---|---|---|
| Autorouter | 158 | 145 | 194.4455 | ~6 |
| Toporouter (before optimization) | 158 | 158 | 190.8683 | 44.966 |
| Toporouter (detour optimization) | 158 | 158 | 188.7315 | 42.496 |
| Toporouter (best optimization) | 158 | 158 | 160.0793 | 51.541 |
| Total Nets | Completed Nets | Wiring Length (inches) | Runtime (seconds) | |
|---|---|---|---|---|
| Autorouter | 11 | 11 | 9.9016 | ~2 |
| Toporouter (before optimization) | 11 | 11 | 8.9696 | 0.332 |
| Toporouter (detour optimization) | 11 | 11 | 8.3151 | 0.424 |