https://open-source-silicon.dev logo
#openlane
Title
# openlane
d

dragonmux

10/22/2021, 11:30 PM
we're trying with sky130_fd_sc_hs as when we try with the sky130_fd_sc_hd cells it bombs out way earlier (step 8 ) with a segfault
v

Vijayan Krishnan

10/23/2021, 5:46 AM
can you share your example? with complete log. If we're able to reproduce will find a solution
d

dragonmux

10/23/2021, 12:12 PM
We can, there are two options - either a link to the source nMigen and you set up your own venv with the patched tools (they're all on GitHub) that generate input for the flow and then execute the flow, or we tar the build files from nMigen that serve as input to OpenLANE. Which is your preference?
v

Vijayan Krishnan

10/23/2021, 12:15 PM
yes just provide tar file that input to Openlane
d

dragonmux

10/23/2021, 12:23 PM
👍 see attached. There are two wrapper scripts in the archive (one for windows and one for everything else) which you can run to initiate the flow how we did. You'll need to adjust the paths to point to your own copy of OpenLANE and the PDKs, and map the directory you extract the file into into the OpenLANE container but that should just work to repro this.
by changing the config.tcl to use sky130_fd_sc_hs instead of hd in this, you should then be able to go on and repro our other flow issue which has us stumped
v

Vijayan Krishnan

10/23/2021, 12:38 PM
are you trying caravel platform with MPW-3 tag OL version?
d

dragonmux

10/23/2021, 12:39 PM
didn't see an MPW-3 tag, currently this is being built with OpenROAD project's latest OpenLANE tag (2021.10.08_02.00.18)
which matches the published container
v

Vijayan Krishnan

10/23/2021, 12:40 PM
ok
d

dragonmux

10/24/2021, 7:15 PM
so with the diodes problem, fixed it by changing strategies - 2 seems broken, but 4 (Sylvain's custom scripts) worked flawlessly and after fixing a pre-OpenLANE issue in nMigen to do with the clock tree stuff, we now have GDS's falling out of 'No Fatal Errors' runs