open-source-silicon.dev

Does anyone know what the klayout_poly_density check in the efabless tapeout job check for? Is there a lower limit of pl_target_density we need to set for this?

In general the low end will be taken care of for you by fill added automatically by efabless.

I got multiple DRC violations for klayout_poly_density check on efabless, that's why I'm wondering whether there is a limit?

Ah you're probably hitting the maximum of 38%

That's because the decap_12 cells have too much poly in them. There was an update to open_pdk to try and use some othr cells to fill things up with less poly density.

Oh ok, but it's better to use the default pdk version for efabless tapeouts right? Instead of a different version? I'll try to edit the config file for the correct decap cell for now

<@U07CB957B54>:  This is a new/tighter rule requirement that was recently imposed by SkyWater, which is why the problem is showing up now.  Unfortunately, the tighter rule for poly density requires less poly density than is in the decap_12 cell, so any design which is sufficiently large and sparsely populated with logic will now fail poly density checks, where it would have passed previously.  We have found a workaround which is to change the settings in openlane to use decap_3 as fill instead of decap_12.  It has the unfortunate side effect of growing the file size of the project because the number of fill cells increases by a factor of 4, but it does avoid the poly density check failure.