Closed smunaut closed 2 months ago
Does it make more sense to just fix the sky130_fd_sc_hd
version of the cell (which avoids the .lib file issue)?
Has any check been done to make sure that the fixed version correctly abuts all other cells in all positions/orientations?
Fixing sky130_fd_sc_hd
definitely can make sense, I was just under the impression that the _fd_
cells had to remain exactly what we got from skywater.
I did check that the fixed version should have a npc
layer compatible with most the other cells in the library ( Wrote some gdspy
script that was checking that ). The only cell I found to be incompatible is and2_0
. It's incompatible both before and after the changes, that cell also has weird npc
geometry which makes it incompatible with a lot of other cells.
I also patched the cell in a few designs we had in tiny tapeout that were exibiting the issue and that fixed it without causing any new DRC errors.
I didn't check any other layers ( nsdm
/ psdm
/ ... ) because I didn't touch those and we didn't have any reports of errors with those.
@smunaut : I do not like to make changes to the sky130 PDK unnecessarily, but this is a long-standing error that I am well aware of and would like to have a patch for in the upstream repo. A fix to and2_0
would be helpful as well.
A fix as a PR to https://github.com/efabless/skywater-pdk-libs-sky130_fd_sc_hd is preferred.
Ack, I'll see what I can do for and2_0
and open a PR against the above repo instead.
The current conb cell has an NPC layer that's not entirely compatible with other cells in the library and can cause DRC failure depending on what's put left / right of it.
Draft PR as this currently doesn't include any '.lib' files that are needed so flow ends with