Q: Is there some kind of objective measure of the routability of a design?

[donn]

I'm running an experiment (#10176 related) and I'm wondering if there's some kind of like value that would be an objective measure of the routability of a design. Right now GRT seems to be kind of binary, reporting "congestion" around whether there is GRT overflow or not, but is there a way to know if routability is affected even for designs that do not have GRT overflow?

[maliberty]

We have the congestion heatmap you can look at (or dump as a csv). You might invent a metric like number of gcells > 80% utilized but there isn't an "official" one.

[donn]

Thanks for the quick response!

[donn]

For peeps googling in the future, to export a heatmap as a csv, OpenROAD must be running in -gui mode and:

gui::set_heatmap "Routing" "Type" "Congestion"
gui::set_heatmap "Routing" rebuild
gui::dump_heatmap "Routing" congestion.csv

[maliberty]

Another metric is the number of grt or drt routing iterations (less is better)