I've noticed some differences between running the

opensta

aka

sta

standalone executable vs using the STA commands inside the openroad executable. Specifically, with standalone

sta

, you can get a longest delay path using just the mapped netlist Verilog plus the Liberty file. But when using the openroad executable, it seems you must read both the tech lef and the cell lefs in order for the STA commands to do anything useful. Is this expected behavior?

yes, is this a problem?

Hi Matt, I didn't see this until now. It just confused me and is seemingly inconsistent (my perhaps naive thinking: you must not really need the tlef/lefs if standalone sta can do the computation without them, so why do the sta commands in openroad require them?)

in standalone sta the netlist is built from just .v/.lib. In OR we use odb which is a physical database and its primitives are LEF macros. We could support a timing only flow with some work but standalone sta does that just fine

Thanks, that makes sense!