A git pull of siliconcompiler ? I've done it and it seems that it doesn't fix the issue.
You can see an updated version of my PDK file on the link in the original post (I've updated it after pulling master version of SC) so now I have :

# APR tech filefortoolin ('openroad','klayout','magic'):chip.set('pdk','aprtech',tool,stackup,libtype,'lef',os.path.join(sclib_dir_ref,"techlef",f"{sclib_name}.tlef"))

for SKY130 which should match the content of skywater130 which is :

# Tech filefortoolin ('openroad','klayout','magic'):chip.set('pdk','aprtech',tool,stackup,libtype,'lef',pdkdir+'/apr/sky130_fd_sc_hd.tlef')

I don't really know about techlef format thoug, it comes from SKY130 master too.

Yes, "a git pull from siliconcompiler." Now that you are up to date, have you done a diff between the tlef that we have included in our repo?
https://github.com/siliconcompiler/siliconcompiler/blob/main/third_party/pdks/skywater/skywater130/pdk/v0_0_2/apr/sky130_fd_sc_hd.tlef

Another possibility, are you running this in an active openroad prompt.
If so you may need to quit out of it to "delete the design", otherwise errors like "library already loaded" would pop up.

Hard to debug like this, so just offering possible some possibilities...

I'm running it from siliconcompiler only, so no problem on this end.
I'll check the tlef diff, but please know that your help is greatly appreciated nonetheless !

So, I did the diff and the only difference is that my version got this addition :

LAYER nwell  TYPE MASTERSLICE ;  PROPERTY LEF58_TYPE "TYPE NWELL ;" ;END nwellLAYER pwell  TYPE MASTERSLICE ;  PROPERTY LEF58_TYPE "TYPE PWELL ;" ;END pwell

Which should probably not cause any major issue at this point.

By the way, I getFOREIGN statements in the lef file, which seems to cause issues too... Where those removed in siliconcompiler versions ?

On this subject, it seems that the.lyt file must directly reference the.lef file used by the design.
The lef file passed through silliconcompiler have no impact with regard to this issue.

Moreover, if the file path is relative, it must start with./ even if the path following starts with../../. Seesky130A_hd.lyt.txt for details.

Thank you! I think it would also be helpful to see copies of the following files:

• sky130_fd_sc_hd.lef
• sky130_fd_sc_hd.merged.cleaned.lef
• sky130_fd_sc_hd.merged.lef
• sky130_fd_sc_hd.tlef

My best guess right now is that this is something subtle to do with differences in LEF content. I don't believe OpenROAD needs the tlef and library data merged, rather I think the "merged" in the filename in our case refers to a merged file containing all std cell definitions. We took our LEF files from theOpenROAD-flow-scripts repo, rather than directly from the open PDKs repo (so we were not responsible for the preprocessing).

I'll send them to you, just for information, merged is the merging of the .tlef + the .lef using mergeLef tool from the openroad utils. The cleaned was an unsuccessful attempt and is not meaningful.

Please find all mentioned files in this archive :sky130_fd_sc_hd_lef.tar.gz

@suzizecat I am hoping this is a dead thread after 3 years, but just checking. A lot has happened with SC since 2022.

Hi@aolofsson , It is indeed a dead thread (mainly because it was an experiment more than anything, so I went on other experiments) but IIRC I never quite fixed those issues in the end.

I'll give another spin to SC when I'll have a suitable project to play around with 🙂