<@U016EM8L91B> I was thinking that maybe it would be possible to change netgen to ignore disconnecte...
m
@User I was thinking that maybe it would be possible to change netgen to ignore disconnected ports for a circuit as long as there were devices in that circuit (i.e., not black boxes). Do you see any possibility of missing true errors with such a change? Another possibility would be to not make proxy ports for disconnected ports in circuits that contain devices.
t
Netgen does ignore disconnected ports for a circuit as long as they do not alter the netlist.
There is behavior defined around the "-blackbox" option as to whether subcircuits without any contents should be treated as black boxes or as empty circuits. Both are possible in a netlist, which makes handling it tricky. I'm not sure that the "-blackbox" option works in any meaningful way.
m
@Tim Edwards Thanks for the response. I was referring to this problem where it appears that disconnected ports in instances are also ports in the parent circuit. https://skywater-pdk.slack.com/archives/C017UA7LEUV/p1628746152015600
t
Did I get a reproducible example of that from you?
m
Let me check... I logged an issue at https://github.com/RTimothyEdwards/netgen/issues/29 that has sample data. The good LVS run takes over an hour, though. It was a while ago and I believe you've made some changes to magic and maybe netgen since then. I think both programs were at version 196.
t
That issue, I tracked down to an improper scaling when the "extract style" command is run. Scaling is fine when the default style is used, and then gets screwed up as soon as the "extract style" command is run, no matter what the style (the style doesn't need to be changed; just running the command screws up the scalefactors). I fixed that problem, so the error raised in issue #29 should not occur any more.
m
@Tim Edwards That great! I'll check it out.