https://open-source-silicon.dev logo
Join Slack
Powered by
It's definitely not plug & play.
# caravel
t
It's definitely not plug & play.
a
Are your main issues in hardening your design?
t
Main remaining issues are : • Some DRC errors : A couple are reals and I can manually fix them, annoying but not blocking. I "think" the rest are false positives but I'll need confirmation. Gotta take a screenshot and post in #magic for confirmation. But somehow I don't get the same DRC errors when loading magic interactively with the GUI than in the flow. • Antenna errors : Non zero ... And somehow magic and OR checks seems to not agree. Given the different algorithm, I'd expect some difference but seems I almost always get no overlap between what one found and what the other finds.
(Well there is also the fact I'm using OpenRAM blocks with no timing info so 🤞 )
a
@tnt: My recommendation is to keep everything easily reproducible in case changes happen (unless your hardening your sub-blocks, so it makes since to get them clean anyway).
I can help with examples on fixing the rest of the antenna violations (usually a handful of them, right?).
t
Yeah I'm trying to keep everything reproductible ... I have a custom 
interactive.tcl
because I need to insert some custom OpenDB scripts to fix things up a bit at some points in the flow.
👍 1
With 
DIODE_STRATEGY
1
I get 16 reported antenna by magic and 17 by OpenROAD.
I have a run in progress for strategy 
3
, to see what that gives.
So with 
3
, I get ~24 reported by OpenROAD and 77 by magic.
I also get 
unmatched nets = 1
during LVS and a DRC error explosion with 217216 ?!?
a
You can try strategy 2 now as well if you update open_pdks; should give results similar to 1.
The LVS mismatch only happens with OpenROAD's antenna avoidance flow?
t
@Ahmed Ghazy yes LVS mismatch is only with OR.
a
@tnt: Sometimes I thought I noticed it disconnecting some nets and not re-connecting them back, but I am not sure. @mehdi: any idea what might be happening here?
t
It's like .. VPB and VNB got swapped
Look at the bottom of https://paste.c-net.org/PracticePokey (it's the LVS log)
took me some time to see what was different between the side and then ... in one VNB/VGND and VPB/VPWR are together. And in the other side it's VPB/VGND and VNB/VPWR.
3 Views
Open in Slack
PreviousNext
Powered by