What is exit status 127?

Forums › Problems and Bug Reports

joe areeda
joe areeda
Joined: 13 Dec 10
Posts: 285
Credit: 320378898
RAC: 0
4 Sep 2012 2:05:33 UTC
Topic 196510
(moderation:
    )

    Please see tasks:
    http://einsteinathome.org/task/306409726
    http://einsteinathome.org/task/306410082
    http://einsteinathome.org/task/306410025

    Very little information in the results but I'm sure it's something silly on my part.

    I'm building a department server based on Scientific Linux 6.3. Most of the components have been running for a month or so while I was travelling but the temps were higher than I liked and wanted more memory and disks.

    So I got a full tower with lots of fans, moved everything over there added an SSD and velociraptor, upgraded ram to 32GB and reinstalled the OS.

    I have to build boinc from sources for SL6 but it seems to be working for CPU jobs but not GPU jobs. It is the same GTX 550ti from the previous enclosure.

    I had a similar problem with a different system that fixed itself. That one seemed to be a missing dynamic library but this one gives no details.

    Any ideas?

    joe areeda
    joe areeda
    Joined: 13 Dec 10
    Posts: 285
    Credit: 320378898
    RAC: 0

    What is exit status 127?

    4 Sep 2012 3:51:50 UTC
    Message 111261
    (moderation:
      )

      Well to see if there is a problem with the GPU I downloaded the CUDA SDK and built the samples.

      The deviceQuery program works and the bandwidth test works and is pretty interesting

      Device 0: GeForce GTX 550 Ti
 Quick Mode

      Host to Device Bandwidth, 1 Device(s), Paged memory
      Transfer Size (Bytes) Bandwidth(MB/s)
      33554432 6068.5

      Device to Host Bandwidth, 1 Device(s), Paged memory
      Transfer Size (Bytes) Bandwidth(MB/s)
      33554432 5783.6

      Device to Device Bandwidth, 1 Device(s)
      Transfer Size (Bytes) Bandwidth(MB/s)
      33554432 74495.1

      [bandwidthTest] test results...
      PASSED

      I seem to have errored out of my limit of GPU tasks for the day so, I'll get some sleep and see if the things I needed to get the SDK to compile and run fixed the BRP problems.

      Joe

      joe areeda
      joe areeda
      Joined: 13 Dec 10
      Posts: 285
      Credit: 320378898
      RAC: 0

      Well even though I can run

      4 Sep 2012 17:30:46 UTC
      Message 111262
      (moderation:
        )

        Well even though I can run all the samples in the CUDA SDK I'm still getting an error 127 with no further information.

        I also found out my other systems have crappy transfer rates which is probably why the 550ti is so close to the 560ti in credits. If I can get this system to work I'll try moving the 560ti over and see what happens.

        Anybody have any clues on error 127?

        Jord
        Joined: 26 Jan 05
        Posts: 2952
        Credit: 5893653
        RAC: 204

        Doing some Google searches,

        4 Sep 2012 20:27:15 UTC
        Message 111263 in response to message 111262
        (moderation:
          )

          Doing some Google searches, it appears it's a missing library.
          You could do ldd on the einstein binaries, just to check if there's anything they need that you're missing.

          joe areeda
          joe areeda
          Joined: 13 Dec 10
          Posts: 285
          Credit: 320378898
          RAC: 0

          Thanks Jord! It never

          5 Sep 2012 0:45:12 UTC
          Message 111264 in response to message 111263
          (moderation:
            )

            Thanks Jord!
            It never occurred to me that this wasn't an application specific error.

            You're right ldd shows 

            $ ldd einsteinbinary_BRP4_1.28_i686-pc-linux-gnu__BRP4cuda32nv270
	linux-gate.so.1 =>  (0x00b8e000)
	libcufft.so.3 => not found
	libcudart.so.3 => not found
	libcuda.so.1 => /usr/lib/nvidia/libcuda.so.1 (0x00b8f000)
	libpthread.so.0 => /lib/libpthread.so.0 (0x00369000)
	libm.so.6 => /lib/libm.so.6 (0x00452000)
	libstdc++.so.6 => not found
	libc.so.6 => /lib/libc.so.6 (0x00526000)
	/lib/ld-linux.so.2 (0x00504000)
	libz.so.1 => /lib/libz.so.1 (0x003b4000)
	libdl.so.2 => /lib/libdl.so.2 (0x0090e000)
	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00244000)

            now to see if I can track down which packages have those libraries. I'll report back when I find them.

            Thanks again!

            Joe

            Khangollo
            Khangollo
            Joined: 17 Feb 11
            Posts: 42
            Credit: 928047659
            RAC: 0

            libstdc++.so.6 => not

            5 Sep 2012 0:59:15 UTC
            Message 111265
            (moderation:
              )

              libstdc++.so.6 => not found is the culprit. libcufft and libcudart are automatically downloaded from E@H and both .so files should be in E@H directory (in boinc data dir / projects).

              Look for compat-libstdc++ (32 bit, .i686) packages.

              MaU38.gif

              joe areeda
              joe areeda
              Joined: 13 Dec 10
              Posts: 285
              Credit: 320378898
              RAC: 0

              Well it seems the only

              5 Sep 2012 1:05:15 UTC
              Message 111266
              (moderation:
                )

                Well it seems the only library really missing is the 32bit version of libstdc++

                Fixed with
                yum install libstdc++.i686

                The libcu* are in the projects directory and evidently the application is smart enough to load them explicitly:

                $ ll lib*
-rwxr-xr-x 1 joe joe   337272 Sep  4 17:42 libcudart32_32_16.so
-rwxr-xr-x 1 joe joe 28247088 Sep  4 17:42 libcufft32_32_16.so

                At least BRP4cuda apps now start and run for a least 10 minutes.

                Joe

                joe areeda
                joe areeda
                Joined: 13 Dec 10
                Posts: 285
                Credit: 320378898
                RAC: 0

                RE: libstdc++.so.6 => not

                5 Sep 2012 1:07:26 UTC
                Message 111267 in response to message 111265
                (moderation:
                  )

                  Quote:

                  libstdc++.so.6 => not found
                  is the culprit. libcufft and libcudart are automatically downloaded from E@H and both .so files should be in E@H directory (in boinc data dir / projects).

                  Look for compat-libstdc++ (32 bit, .i686) packages.

                  That's what had me confused. I did have compat-libstdc++-296.i686 installed which is evidently enough to remove the "missing libstdc++" from the output but not good enough to let them run.

                  I'm learning, slowly.

                  Joe

                  joe areeda
                  joe areeda
                  Joined: 13 Dec 10
                  Posts: 285
                  Credit: 320378898
                  RAC: 0

                  that seems to do it. 2 WUs

                  5 Sep 2012 1:50:12 UTC
                  Message 111268
                  (moderation:
                    )

                    that seems to do it.

                    2 WUs completed and validated:

                    http://einsteinathome.org/task/306575855
                    http://einsteinathome.org/task/306575854

                    My thanks to Jord and Khangollo for the help!

                    Joe

                    Comment viewing options

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

                    Forums › Problems and Bug Reports