2 Replies Latest reply on Feb 23, 2019 8:42 AM by growingmind

    Error with Calibre PEX using IBM PDK




      This is an update from my previous message in this community:



      I am now using an updated version of Cadence (1.6.7.b where b = base version) and an updated version of Calibre (2015.2).


      I am still getting errors with the PEX Rules file compilation but no longer with the LVS or DRC rules file compilations.


      Calibre will not compile the PEX rules file and gives me the following error (attached).


      The vendor of the PDK tested this with Cadence IC and Calibre 2015.4.


      A colleague of mine can run the same PDK with Cadence and Calibre 2014, not Calibre 2015.


      What reason, then could it be that I get a compilation error when my Calibre version is newer ?  Is it the older version of Cadence 1.6.7 ?


      Does that affect the running of the Calibre PEX or Calibre LVS or Calibre DRC ?

      The old Cadence version ?


      Can this error be fixed ?  What can I do ?  Contact the PDK Vendor ?  Cadence told me it could not help.


      Thank you.

        • 1. Re:  Error with Calibre PEX using IBM PDK

          Was that the only error message?  It could be that there is a problem with the encryption, or that there is a problem with the SVRF.  If the PDK was developed with Calibre 2015.4 in mind, either could happen with you using an earlier version.  Here are some of the things in the 2015.3 and 2015.4 release notes that could have caused it:

          • Regarding encryption, 2015.3 adds this note:
            Older versions of caltvfencrypt produced output that will not be readable in future versions of Calibre that use newer Tcl interpreters. You may need to re-encode any files that were encrypted using older versions of caltvfencrypt with the current version in order to ensure forward compatibility. Be sure to maintain the original Tcl source code and keep it available for re-encoding with future versions of the encrypter in case of future command or Tcl syntax changes.
            (The note does say "future versions" so it might not have kicked in when the PDK provider encrypted. Best to check with them.)
          • For SVRF changes, there are 6 new PEX SVRF commands (PEX Netlist Comment, Node Prefix, and Short Toplevel Ports; PEX Density Actual and Context; PEX XACT Noxref) and existing statements with new options (PEX Fieldsolver Cell_array adds Z, Positive, Negative, and Ground; PEX Netlist adds LeadingX, Stringquotes, and NetName; PEX Density Window adds Step and Context) and new capabilities (PEX Netlist Auto_LPE can do CalibreView, and several other options now allow wildcards that did not formerly).  If any of those got used, they'd cause a parser error in v2015.2.


          Your best solution is to check with the PDK provider (IBM?) and see if they know of incompatibilities.  They may have a pre-2015.4 version of the PEX rules that will work for you. 


          There have also been many bug fixes and accuracy improvements over the past 4 years; if you can convince your employer it is probably worthwhile to go to 2019.1, or at least 2015.4, as the PDK was qualified with it.


          Good luck-



          • 2. Re:  Error with Calibre PEX using IBM PDK



            Thank you for your response.


            It seems it is not quite as simple as using the new version of Calibre.


            I have a colleague at another company, albeit running on the same version of UNIX on different processors running a pre Calibre 2015.4 version for this kit, and an IXL version, not the AOI version as required for this UNIX version and it is working, even though the PDK says 2015.4 should be used.  And he is forcing IXL to work on his UNIX version using an environmental variable to do so.


            So the answer is not quite as simple as there seems to be a different way Calibre runs on different  UNIX versions (ie. RedHat 5.x, 6x) and the encryptions run differently between AOI and IXL.  It seems processor dependent or computer dependent and even possibly dependent on the version of Cadence running.


            Any thoughts on this ?