Task 'Running out of paper' !!!

Forums › Problems and Bug Reports

BobMALCS
BobMALCS
Joined: 13 Aug 10
Posts: 20
Credit: 54539336
RAC: 0
29 Dec 2016 21:11:02 UTC
Topic 204174
(moderation:
    )

    I don't know whether to laugh or cry at this one.

     

    Task 598996629

    Name: LATeah0010L_908.0_0_0.0_10782960_0
    Work unit ID: 267609487
    Created: 29 Dec 2016 16:40:23 GMT
    Sent: 29 Dec 2016 17:36:11 GMT
    Received: 29 Dec 2016 21:00:16 GMT
    Server state: Over
    Outcome: Computation error
    Client state: Compute error
    Exit status: 28 (0x0000001C) Unknown error code
    Computer: 6630273
    Report deadline: 12 Jan 2017 17:36:11 GMT
    Run time (sec): 2,487.47
    CPU time (sec): 2,457.66
    Validation state: Invalid
    Claimed credit: 25.73
    Granted credit: 0.00
    Application: Gamma-ray pulsar binary search #1 on GPUs v1.17 (FGRPopencl-nvidia) windows_x86_64

    Stderr output

    <core_client_version>7.6.33</core_client_version>
<![CDATA[
<message>
The printer is out of paper.
 (0x1c) - exit code 28 (0x1c)
</message>
<stderr_txt>

    .
    .
    .
    .
    .
    .
    .
    % Binary point 12/1255
    % Starting semicoherent search over f0 and f1.
    % nf1dots: 31 df1dot: 3.344368011e-015 f1dot_start: -1e-013 f1dot_band: 1e-013
    % Filling array of photon pairs
    .
    .

     

    Lots of '.' and then it starts Binary point 13/1255 and so on.  Eventually it runs out of paper.

     

    Any thoughts ?

     

    BobM

    Gary Roberts
    Gary Roberts
    Moderator
    Joined: 9 Feb 05
    Posts: 5849
    Credit: 110009666549
    RAC: 24059297

    Of course if you scroll right

    29 Dec 2016 23:36:11 UTC
    Message 153515
    (moderation:
      )

      Of course if you scroll right to the bottom of the stderr output you will find

      ERROR: /home/bema/fermilat/src/bridge_fft_clfft.c:930: clFinish failed. status=-36
ERROR: opencl_prepare_power_toplist() returned with error 7338939
20:54:28 (10412): [CRITICAL]: ERROR: MAIN() returned with error '-36'
FPU status flags: PRECISION
20:54:40 (10412): [normal]: done. calling boinc_finish(28).
20:54:40 (10412): called boinc_finish

      which is probably a much better indication of what the problem really was.

      Cheers,
      Gary.

      BobMALCS
      BobMALCS
      Joined: 13 Aug 10
      Posts: 20
      Credit: 54539336
      RAC: 0

      Ok.  But what is the

      30 Dec 2016 14:18:05 UTC
      Message 153553 in response to message 153515
      (moderation:
        )

        Ok.  But what is the implication of 'FPU status flags: PRECISION'

        Program error, bad calculation, GPU problem, ...

        BobM

        archae86
        archae86
        Joined: 6 Dec 05
        Posts: 3145
        Credit: 7059424931
        RAC: 1263442

        I've been getting a higher

        30 Dec 2016 15:13:36 UTC
        Message 153555
        (moderation:
          )

          I've been getting a higher rate of computation error failures in my fleet on the FGRBP1 work than for the last year running the previous two primary Einstein GPU applications.  It is well under 1% and scattered across my machines.  I've not tried reducing overclocks to try to find whether what was a safe overclock for the previous work is just a little too high for this one.

          One unifying theme I have seen is seemingly spurious error messages, as reported in the starting post of this thread.  In fact I, too, got an out of paper message on at least one computation error stderr.

          I have also gotten

          The network BIOS session limit was exceeded.
 (0x45) - exit code 69 (0x45)"

          The network name cannot be found.
 (0x43) - exit code 67 (0x43)

           

          Logforme
          Logforme
          Joined: 13 Aug 10
          Posts: 332
          Credit: 1714373961
          RAC: 0

          archae86 wrote:One unifying

          30 Dec 2016 15:38:33 UTC
          Message 153556 in response to message 153555
          (moderation:
            )

            archae86 wrote:
            One unifying theme I have seen is seemingly spurious error messages, as reported in the starting post of this thread.  In fact I, too, got an out of paper message on at least one computation error stderr.

            I would guess that these error messages is a result of the program asking windows for the error text associated with an error code. Only problem is that the error code comes from opencl (or internal codes from the program), not windows, so you get a nonsense text.

            Windows error codes

            Gary Roberts
            Gary Roberts
            Moderator
            Joined: 9 Feb 05
            Posts: 5849
            Credit: 110009666549
            RAC: 24059297

            BobmALCS wrote:Ok.  But what

            31 Dec 2016 0:14:01 UTC
            Message 153571 in response to message 153553
            (moderation:
              )

              BobmALCS wrote:
              Ok.  But what is the implication of 'FPU status flags: PRECISION'

              It's not part of the error message - it's there in the stderr output on 'good' results as well as 'bad'.

              It's listed as a FPU status flag.  Maybe "PRECISION" just means the FPU is giving the best possible precision for the calculations it performs.  That would be my guess.

               

              Cheers,
              Gary.

              BobMALCS
              BobMALCS
              Joined: 13 Aug 10
              Posts: 20
              Credit: 54539336
              RAC: 0

              Gary Roberts wrote:Of course

              2 Jan 2017 13:35:26 UTC
              Message 153668 in response to message 153515
              (moderation:
                )

                Gary Roberts wrote:

                Of course if you scroll right to the bottom of the stderr output you will find

                ERROR: /home/bema/fermilat/src/bridge_fft_clfft.c:930: clFinish failed. status=-36
ERROR: opencl_prepare_power_toplist() returned with error 7338939
20:54:28 (10412): [CRITICAL]: ERROR: MAIN() returned with error '-36'
FPU status flags: PRECISION
20:54:40 (10412): [normal]: done. calling boinc_finish(28).
20:54:40 (10412): called boinc_finish

                which is probably a much better indication of what the problem really was.

                I have no doubt that someone knows what the error return means but I don't. 

                So until there is some indication of what the problem is and possibly a solution I wont be running Einstein.

                BobM

                Christian Beer
                Christian Beer
                Joined: 9 Feb 05
                Posts: 595
                Credit: 127321528
                RAC: 354165

                BOINC on windows tries to

                2 Jan 2017 14:08:42 UTC
                Message 153670
                (moderation:
                  )

                  BOINC on windows tries to guess the human readable version of an error code and interprets the exit code of an app as if it where a windows error code. This is not the case here.

                  As seen in the stderr.log the real error code is -36 which is openCL specific and translates to CL_INVALID_COMMAND_QUEUE which is something Bernd will have to look at when he is back.

                  BobMALCS
                  BobMALCS
                  Joined: 13 Aug 10
                  Posts: 20
                  Credit: 54539336
                  RAC: 0

                  Christian, thanks for the

                  2 Jan 2017 22:01:25 UTC
                  Message 153690 in response to message 153670
                  (moderation:
                    )

                    Christian, thanks for the info.    I'll wait patiently.

                    BobM

                    Comment viewing options

                    Select your preferred way to display the comments and click "Save settings" to activate your changes.

                    Forums › Problems and Bug Reports