wafer.space Community - False positive DRC PP5.a andV1.3c

See the attached design. AFAICT the PP5.a errors and V1.3c errors are false positives ...

70.94 KB

157.76 KB

https://gf180mcu-pdk.readthedocs.io/en/latest/physical_verification/design_manual/drm_07_15.html V1.3c is the Vn.3 (iii) clause, right?

I'm not sure because the limit is 0.06 ... and (iii) looks more like a recommendation ...

6:08 p.m.

I think it's (ii) because there is actually a rule before (i)

oh right forgot that the plain one also ought to count as a sub-rule

6:18 p.m.

Though I can't see how (i) isn't superceeded by (ii), but oh well...

Me either ... and TBH I can't make heads or tails of the rule implementation in the deck so no clue how to fix it ... it's obviously confused by the nearby 45deg corner but it's still a valid enclosure and shouldn't throw a DRC error.

6:24 p.m.

(and it doesn't even throw one at all such corner so obviously the rule is not even deterministic since the geometry orientation affects the result showing it's a bad rule implementation). (edited)

@tnt Could you please open an issue at: https://github.com/fossi-foundation/globalfoundries-pdk-libs-gf180mcu_fd_pv I can probably take a closer look by the middle of next week.

GitHub - fossi-foundation/globalfoundries-pdk-libs-gf180mcu_fd_pv

Contribute to fossi-foundation/globalfoundries-pdk-libs-gf180mcu_fd_pv development by creating an account on GitHub.

PP5.a can be worked around by using run_mode=flat ... but looks to me like the deep mode is completely broken because it doesn't even merge polygons at the same hierarchy level.

6:42 p.m.

@Leo Moser (mole99) Will do

Some operations perform merging under the hood, while for others you either don't want due to performance reasons, or need to merge explicitly. https://www.klayout.de/drc_function_internals.html

ATM there is merge = (ctx.options.run_mode != 'deep') So if run_mode is anything but deep it performs a merge on the layer it extracts. (edited)

6:51 p.m.

ATM on a test design :

deep : 12 sec but wrong results
flat: 242 sec but corrects results
deep but with merge = true : 153 sec and different wrong results ...

6:52 p.m.

Ok, gtg now, but I'll open an issue tonight with example files.

Mmm ... part of it could actually be a KLayout bug ... Seems not to reproduce on my build VM. I'm updating my local klayout to check if that's the difference.

Ok, seems like the PP5.a are fixed by KLayout update. Doesn't fix the V1.3c though and those are not fixed by flat mode or anything.