| # Detailed routing time for mock-array as a function of area |
|
|
| Tool: Detailed Routing |
|
|
| Subcategory: Performance analysis |
|
|
| ## Conversation |
|
|
| ### oharboe |
| Using https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/pull/1211 and https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/pull/1212, I've plotted detailed routing times for the same Verilog 8x8 8 bit datapath mock-array, changing only the size of the element macros.
|
|
|
| Any thoughts on these running times for the mock-array for detailed routing? Expected? Unexpected?
|
|
|
| The algorithm seems to be MUCH faster for smaller areas
|
|
|
| 
|
|
|
|
|
| Biggest array:
|
|
|
| 
|
|
|
| The element-element wires are not straight, lots of little deviations from horizontal/vertical:
|
|
|
| 
|
|
|
| Smallest array:
|
|
|
| 
|
|
|
|
|
| "All" element to element wires are horizontal/vertical (didn't check exhaustively):
|
|
|
| 
|
|
|
|
|
|
|
| ### maliberty |
| How do the number of routing iterations and iter0 DRCs compare between datapoints? |
|
|
| ### oharboe |
| I wasn't able to follow up on this in a timely manner. |
|
|
|
|