4 Replies Latest reply on Mar 1, 2018 1:57 PM by sam_1980

    Calibre PEX violations




      We have recently upgraded to Calibre release "v2013.2_18.13" and started getting the following errors which were not there before.


      WARNING:  Hcells "nmos_rf_25*" and "nmos_rf_25" are of different kinds  ("Subcircuit") and ("LVS box or empty subcircuit") - correspondence  ignored.


      Hcell correspondence "nmos_rf_25*" "nmos_rf_25" leads to a many-many correspondence.

      Corresponding cells could not be identified.

      Correspondence "nmos_rf_25*" "nmos_rf_25" leads to a many-many correspondence.



      All the files remain the same.


      Any ideas would definitely help!!!


      I was able to run the standalone Hierarchical LVS without any problems.



        • 1. Re: Calibre PEX violations

          If I include LVS EXCLUDE HCELL "nmos_rf_*" in the PEX options-->Include, then it's working fine...Not sure why we didn't have to write this extra with our old Calibre version or why I have to include this in the newer Calibre release.



          • 2. Re: Calibre PEX violations

            Hi --


            If you were updating from 2009.3, which was a version you mentioned a few months ago on the forum, I suspect the cause was that in 2009.4 Calibre started supporting wildcards for layout names in the Hcell statement.


            Here's the relevant additional information from the 2009.4 Release Notes (p.36):


            Calibre nmLVS does not allow many-to-many hcell correspondence. This restriction remains in

            place when wildcards are used for hcell names. Using wildcards in hcell specifications may be

            included for -spice (extraction) runs in a future release, but current support is for the comparison

            module only. See the SVRF Manual for more details on the restrictions of using wildcards in

            hcell specifications.


            (A better place would be the "Hcell Correspondence" topic in the Calibre Verification User's Manual.)  So that would explain the many-many warning appearing.


            For the "cells are of different kinds" warning, there's a tiny bit more info in the SVRF Manual, in Table 7-73.  Things that count as "different kinds" are cell vs primitive device, or having an empty subcircuit or LVS Box cell as one side of the correspondence. (This seems to be the case from the transcript.)  You would have to look at your source or SVRF file to see which one applies to nmos_rf_25.  I don't know when this check was added, but presumably between the old and new versions.


            What puzzles me is that you were able to dismiss these LVS warnings by setting a PEX option.  Are your LVS and PEX runs using different settings?  (This is usually not a good idea - makes debugging hard.)


            Hope this helps-




            • 3. Re: Calibre PEX violations

              Hi Sam,


              Thanks for the detailed explanation.


              We mostly run "Flat" LVS and only use "Hierarchical" option occasionally (to double-check with the PEX LVS results).


              I will check the info you mentioned.



              • 4. Re: Calibre PEX violations

                Hello Mugofgold,
                I have this problem for mom cap ... could you please let me know what you added to PEX_options? what is the format for rule files like LVS EXCLUDE HCELL "nmos_rf_*"


                Thanks a lot,