AFAICT this is the net with the issue

(at least that's what's connected to instance

_4391_

port

A

)

@User I am not sure what is your question.

You have the instance and net name and port/sink name

If I dump more info we will end up with a huge file

@User I'm just not sure how to read the file ... ie what the various lines mean and how to use that to find the spot where there is an antenna.

The router makes very peculiar choices 🤦‍♂️

@User https://pastebin.com/Zc9LV0G4

@User ^

@User Thanks. But the various listing for the various layers, that doesn't describe the 'net' at all ? Like if the met2 segments are split into several parts etc ?

The OpenRoad and

magic

antenna check seem to find completely different antennas 😕

@User There should be a net name printed at the beginning(with no indentation), then the connected instances are printed.

like this one：

@User also you can see the pin name connected to each net/wire:

_4391_ (instance) (sky130_fd_sc_hd__and3_2)  A (pin/sink)

Magic is doing different (extraction) calculations I think (similar to calibre). While the ARC is using the LEF files to do that

Ok, now of course is how to solve them 😅

Some where easy with ... for some reason the router decided to use

met2

to connect to the diode right next to it instead of

met1

. For others ... the auto-inserted diode is half the design aways 😕

I don't think it is good to use the strategy where you insert a conservative # of diodes

The flow with ARC+FR is good. It is not perfect but it fix most of the issues

the problem you reported is being fixed.

Mmm ... I just tried going back to

DIODE_STRATEGY

3

... and it doesn't do what it used to (AFAICT). However (1) LVS doesn't match anymore and (2) got a bunch of antenna violation at the end.

Oh nevermind, forget what I said, it does ... it just doesn't stop the flow for some reason.

@User we could have a call later today if it helps going through the antenna stuff

I am getting this error, which seems to originate in openroad: runs/user_proj_example/logs/placement/basic_mp.log:[ERROR] Cannot find any macros in this design. (The full log is available here: https://pdk.libresilicon.com/caravel-build.html)

One additional comment: @User I’ve noticed that the script that you’ve sent to me sets the placement padding as 2. I think you don’t need to set padding for diodes, since the minimum filler cell has 1-site width

It should be set like this: set_placement_padding -global -left 0 -right 0

Please try it and let me know how it goes

Can you get PDN gen to not generate rails ? And just straps ?

Mmm ... I'm trying to use the

getAvgXY

API on

ITerms

in the python bindings but it wants

x

and

y

argument ... if I try to feed anything it says it want

int *

which ... is a bit tough in python 😕