climateprediction.net home page

The world's largest climate forecasting experiment for the 21st century.

Hadcm3n models crashing on new Macintosh computers too


Advanced search

Message boards : Number crunching : Hadcm3n models crashing on new Macintosh computers too

AuthorMessage
dajashby
Send message
Joined: Sep 1 04
Posts: 55
Credit: 10,972,690
RAC: 3,387
Message 42431 - Posted 20 Jun 2011 22:39:12 UTC


    hadcm3n is NOT compatible with CPUs/processors without SSE2 capabilities. That means HADCM3N is NOT compatible with


    Pentium 3, Pentium 3 based Celerons and Xeons, and older Intel processors, and
    Athlon XP/MP, Athlon XP based Semprons/Durons, and older AMD processors.



    A computer with a processor listed above downloading one of these models will result in an immediate crash. This is not expected to change in the future.


    geophi, you can add Macintosh computers with Intel's new Sandy Bridge processor to that list. I've now had two of these models crash on my 2011 iMac within seconds of launching - which, given the difficulty of getting any work at all, is pretty frustrating.
    ____________
    Derrick Ashby

    Les Bayliss
    Forum moderator
    Send message
    Joined: Sep 5 04
    Posts: 5128
    Credit: 8,436,525
    RAC: 4,382
    Message 42433 - Posted 20 Jun 2011 23:10:32 UTC - in response to Message 42431.

      Are you sure that's the cause, and not the batch of faulty models as mentioned in recent posts in News and Announcements?

      ____________
      Backups: Here

      Profile Thyme Lawn
      Forum moderator
      Send message
      Joined: Aug 5 04
      Posts: 1212
      Credit: 10,212,294
      RAC: 755
      Message 42434 - Posted 21 Jun 2011 8:05:34 UTC

        The stderr messages for your failed tasks all contain errors similar to the following:

        execl(/Library/Application Support/BOINC Data/projects/climateprediction.net/hadcm3n_um_6.07_i686-apple-darwin, 137095) failed!
        Controller:: CPDN process is not running, exiting, bRetVal = 1, checkPID=0, selfPID=10400, iMonCtr=1
        Model crash detected, will try to restart...

        That computer's task list has a HadAM3P_SAF task crashing with the same error on 26th May and a HadAM3P_EU task completing successfully on 22nd May. The other notable difference between them is that the EU task reported BOINC core client version 6.10.58 and the SAF one reported 6.12.26, so the error could indicate that the upgrade caused problems with the permissions.

        Interestingly, the SAF task returned 5 trickles after the error was reported, suggesting that you managed to get it to run from a backup. How did you manage that?
        ____________
        "The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

        dajashby
        Send message
        Joined: Sep 1 04
        Posts: 55
        Credit: 10,972,690
        RAC: 3,387
        Message 42453 - Posted 23 Jun 2011 22:15:40 UTC


          That computer's task list has a HadAM3P_SAF task crashing with the same error on 26th May and a HadAM3P_EU task completing successfully on 22nd May. The other notable difference between them is that the EU task reported BOINC core client version 6.10.58 and the SAF one reported 6.12.26, so the error could indicate that the upgrade caused problems with the permissions.


          I acquired the current computer (iMac with i7 processor) in late May, but made the mistake of migrating the settings from the laprop (Macbook Air) I had been using (I'm new to the Macintosh). The BOINC settings were transferred along with everything else, and although I reinstalled BOINC it may be that I need to clear the whole thing out and start again. The successfully completed tasks were on the laptop. Everything that's run on this machine from CPDM has crashed.

          The case where trickles arrived after the model crashed was probably because the same model was being run on both machines, which was a bit stupid of me.
          ____________
          Derrick Ashby

          dajashby
          Send message
          Joined: Sep 1 04
          Posts: 55
          Credit: 10,972,690
          RAC: 3,387
          Message 42455 - Posted 24 Jun 2011 13:09:15 UTC

            OK, I have cleared out my BOINC Data folder and reinstalled the BOINC manager, and, somewhat to my surprise, picked up a new hadcm3n model immediately. It has so far run for several minutes without crashing. I'm crossing my fingers.
            ____________
            Derrick Ashby

            dajashby
            Send message
            Joined: Sep 1 04
            Posts: 55
            Credit: 10,972,690
            RAC: 3,387
            Message 42464 - Posted 25 Jun 2011 13:29:34 UTC - in response to Message 42455.

              15 hours now...:-)
              ____________
              Derrick Ashby

              dajashby
              Send message
              Joined: Sep 1 04
              Posts: 55
              Credit: 10,972,690
              RAC: 3,387
              Message 42466 - Posted 26 Jun 2011 2:15:37 UTC - in response to Message 42464.

                And now I have a second hadcm3n model. Looks like it was my BOINC installation causing the problems.
                ____________
                Derrick Ashby

                Post to thread

                Message boards : Number crunching : Hadcm3n models crashing on new Macintosh computers too




                Copyright © 2002-2014 climateprediction.net