| # Running times for 6_report as a function of area | |
| Tool: Global Routing | |
| Subcategory: Performance issue | |
| ## Conversation | |
| ### oharboe | |
| Using https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/pull/1211 and executing `python util/plot-running-times.py`, I've plotted some running times against area: | |
| `6_report` running time in seconds below. Note that the x axis is um^2, but that it is multiplied by 1e6 (as can be seen in the lower right corner). | |
| As can be seen, the running time is worse than proportional to area and I imagine that there could be some low hanging fruit as I would expect reporting not to be on the order of detailed routing times. | |
|  | |
| ``` | |
| cat /proc/cpuinfo | grep processor | wc -l | |
| 12 | |
| ``` | |
| ### oharboe | |
| Same for `2_6_pdn`, here running time is proportional to area: | |
|  | |
| ### oharboe | |
| `5_2_TritonRoute` running times are misleading as there's some sort of problem with DRC errors #3592 | |
| However, since I have the data, I include the plot here anyway. | |
|  | |
| ### oharboe | |
| A feature request has been filed to improve reporting times #3599 | |