open-source-silicon.dev

<@U016EM8L91B> With 1.5.157 ,I now get an error that "The top level cell failed pin matching." and the netlists do not match. Whereas in version 1.5.117 it passed with only a "Cell pin lists for replica_cell_6t and replica_cell_6t altered to match."

<@U016EM8L91B>: Is this somehow related to the issue I reported earlier on our <#C016DMRJA14|skywater>?

<@U016EM8L91B> That command is magic. (pun intended)

readspice.tar.gz

<@U016EM8L91B> Though it fails for some odd reason on this case. The case intentionally has 2 ports that are "disconnected" (not connected to any devices),  but everything else seems like it should match from the output. I have a single script that does extraction and LVS.

Previously, I had to flatten these cells to resolve a symmetry with them (that is in the setup.tcl)

<@U0175T39732>:  Add the "-full" flag after "-json" in your call to LVS.  I changed the behavior of the symmetry breaking to keep standard cell designs from taking hours to complete.  Hopefully some time I will figure out a way to detect which situation requires which handling of symmetry breaking---the digital standard cell problem has to do with cells not connected to other cells like tap cells or antenna tie-downs.  I prefer the way I have it now even though it is not backwards-compatible, because for small cells, the time to run LVS and discover that you need the "-full" switch is small, whereas if I did it the other (backwards compatible) way, then for large digital standard cell designs, the time to run LVS and discover that you should have used the "-fast" switch can be hours and hours.  Automatically choosing is of course preferred, but I haven't gotten there yet.

<@U0175T39732>:  Although I will add that the behavior of netgen reporting pin mismatches after incorrect symmetry breaking is one I had not seen before until today.  <@U016HSALFAN>, though, reported earlier today on an example that he says he ran with "-full" and still resulted in a weird pin match error.  Now I have more problems to solve. . .

<@U016EM8L91B> That option made my unit tests go from 1 hour to more than 3 (it hit my time limit).

hello i have the same error, i add -full before LVS but nothing works for me

<@U03CSQTGFM2>:  That issue was raised three years ago, worked on, resolved, and the "-full" vs. "-fast" flags were deprecated (no longer have any function).  If you're seeing long run times, create a new issue, because I very much doubt it's related to anything discussed above.