Forum Discussion

DerekNH's avatar
DerekNH
Icon for New Contributor rankNew Contributor
2 years ago
Solved

Quartus Prime Pro 23.4 internal error during fit when module marked as a partition

I am working to package my portion of a Stratix 10 design so that it can be used as a "finalized" partition within a larger FPGA implementation. However, when I mark my top level module as a partition, Quartus crashes with an internal error 13% of the way through the fitter place stage. Without the partition, implementation is successful. With a logic lock region for my top level module but no partition, implementation is also successful.

Any suggestions how I can avoid the internal error?

Details:

Quartus Pro v23.4 (also crashes on v23.3 but I upgraded to the latest to check if resolved)
Install-default Quartus settings
Stratix 10, my portion of the design uses ~15% of device resources
Ubuntu 20.04.3 LTS
AMD Ryzen thread-ripper 3970x 32-core
256 GB

Crash log attached

11 Replies

    • DerekNH's avatar
      DerekNH
      Icon for New Contributor rankNew Contributor

      Indeed, Quartus Pro 23.2 is able to complete implementation of my design even when marked as a partition (and placed in a logic lock region). HOWEVER, I am not sure that 23.2 is actually "DNI"-free. I see the message below during my from-RTL synthesis and fit.

      Info: Running Quartus Prime Synthesis
      Info: Version 23.2.0 Build 94 06/14/2023 SC Pro Edition
      Info: Processing started: Fri Mar 1 11:28:22 2024
      Info: System process ID: 978588
      Info: Command: quartus_syn --dni --read_settings_files=on --write_settings_files=off apipe_top -c apipe_top
      Info: The application is running in 'DNI' mode.

      Thx for the suggestion.


      • roeekalinsky's avatar
        roeekalinsky
        Icon for Contributor rankContributor

        Interesting result.

        Re 23.2 and DNI, my previous understanding was that 23.2 included DNI as an option, presumably for testing, but also still included the legacy framework which was used by default unless the tool is invoked with the --dni switch explicitly. Looks like maybe it's not as simple as that. Did your project scripts include the --dni switch, perhaps a carryover from your previous work on the later versions? Or did 23.2 just spontaneously enable DNI?

        Either way, I'm glad you were able to get through the flow in 23.2, so you at least have a working solution in the interim. And thanks for trying it, and for reporting your findings.

        Now hopefully the Intel folks can jump in here, chase down the bug you discovered in 23.3 & 23.4, and get it fixed in a coming release.

  • DerekNH's avatar
    DerekNH
    Icon for New Contributor rankNew Contributor

    Just to follow up ...

    My scripts at this point are quite bare bones so as to run Quartus in its most mainstream (ie. bug free?) configuration. There is nothing in them to steer either toward or away from DNI ...

    Thx,

    Derek

  • SyafieqS's avatar
    SyafieqS
    Icon for Super Contributor rankSuper Contributor

    Derek,


    I need the design if possible to root cause this.

    Seem DNI issue is getting well known.

    Or any small design that replicate the issue.


  • SyafieqS's avatar
    SyafieqS
    Icon for Super Contributor rankSuper Contributor

    Let me know if there is any update from previous reply.


  • DerekNH's avatar
    DerekNH
    Icon for New Contributor rankNew Contributor

    I am unable to provide the code in its current state due to IP concerns but am working to put together a reduced form which still exhibits the issue.

    Thx,

    Derek

  • SyafieqS's avatar
    SyafieqS
    Icon for Super Contributor rankSuper Contributor

    Understood. Is 23.2 earlier is feasible for current development?

    I would suggest to use to as workaround.

    It is hard if there no design to replicate to root cause the issue and would be better if you are able to provide.



    • DerekNH's avatar
      DerekNH
      Icon for New Contributor rankNew Contributor

      Is there a flow by which we can provide a testcase privately and under NDA? I have been unable to simplify the test case in a way which reproduces the error and does not have proprietary code.

      Otherwise, yes, we will use 23.2 as a workaround.

      Thanks so much,

      Derek

  • SyafieqS's avatar
    SyafieqS
    Icon for Super Contributor rankSuper Contributor

    As we do not receive any response from you on the previous question/reply/answer that we have provided. Please login to https://supporttickets.intel.com/, view details of the desire request, and post a feed/response within the next 15 days to allow me to continue to support you. After 15 days, this thread will be transitioned to community support. The community users will be able to help you on your follow-up questions.


    p/s: If any answer from community or Intel support are helpful, please feel free to mark as solution, give Kudos and rate 5/5 survey