[Openmd-users] (no subject)
Dan Gezelter
gezelter at nd.edu
Mon Jul 19 10:21:04 EDT 2010
David,
It would be helpful if you could send a copy of the md and frc files that he is using so that we can attempt to recreate this. Also, does the same job fail on a single processor? (i.e. if the openmd binary is run instead of mpirun -np N openmd_MPI)? If this is happening in the single processor version as well as under MPI, the debugging will be much easier.
The debugging flags can give you more of a verbose backtrace in idb or gdb, so it may help to reconfigure with "--enable-debug". That said, debugging in openmpi is a bit of a challenge, but if you are adept at using CLI debuggers you can start up a parallel debugging job : mpirun -np 4 xterm -e gdb openmd_MPI
and then use "run file.md" in each of the windows to start up the parallel job.
On RHEL 5.5, we've built and tested with the pathscale compilers (version 3.2) and openmpi (version 1.3.2), but I think I can do a test build with the intel compilers if necessary.
Best regards,
--Dan Gezelter
On Jul 19, 2010, at 5:52 AM, Baker D.J. wrote:
> Hello,
>
> We are new to OpenMD, and recently we installed Openmd 1.1 on our linux cluster at the request of one of our users. We don't, at this stage, have many details of what he's doing, however his job aborts with the trace back as shown below. We build Openmd 1.1 with the Intel compilers v11.1 and openmpi v1.3.3 on a rhel 5.3 machine. Do anyone have any advice for us please?
>
> Should we, for example, preferably build with the gnu compilers, and/or another MPI library? Alternatively should we rebuild OpenMP with debugger support?
>
> Best regards -- David Baker.
>
> terminate called after throwing an instance of 'std::bad_cast'
> what(): St8bad_cast
> [red0576:07809] *** Process received signal ***
> [red0576:07809] Signal: Aborted (6)
> [red0576:07809] Signal code: (-6)
> [red0576:07809] [ 0] /lib64/libpthread.so.0 [0x2b13b79684c0]
> [red0576:07809] [ 1] /lib64/libc.so.6(gsignal+0x35) [0x2b13b88e5215]
> [red0576:07809] [ 2] /lib64/libc.so.6(abort+0x110) [0x2b13b88e6cc0]
> [red0576:07809] [ 3] /usr/lib64/libstdc++.so.6(_ZN9__gnu_cxx27__verbose_terminate_handlerEv+0x114) [0x2b13b8465c44]
> [red0576:07809] [ 4] /usr/lib64/libstdc++.so.6 [0x2b13b8463db6]
> [red0576:07809] [ 5] /usr/lib64/libstdc++.so.6 [0x2b13b8463de3]
> [red0576:07809] [ 6] /usr/lib64/libstdc++.so.6 [0x2b13b8463eca]
> [red0576:07809] [ 7] /usr/lib64/libstdc++.so.6 [0x2b13b8400142]
> [red0576:07809] [ 8] /usr/lib64/libstdc++.so.6 [0x2b13b8415e98]
> [red0576:07809] [ 9] /usr/lib64/libstdc++.so.6(_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_+0x15) [0x2b13b8437005]
> [red0576:07809] [10] openmd_MPI(_ZN6OpenMD10StatWriter9writeStatERKNS_5StatsE+0x139) [0x60c8e9]
> [red0576:07809] [11] openmd_MPI(_ZN6OpenMD24VelocityVerletIntegrator8postStepEv+0x15c) [0x59e69c]
> [red0576:07809] [12] openmd_MPI(_ZN6OpenMD24VelocityVerletIntegrator11doIntegrateEv+0x43) [0x59e4a3]
> [red0576:07809] [13] openmd_MPI(main+0xa30) [0x474fd0]
> [red0576:07809] [14] /lib64/libc.so.6(__libc_start_main+0xf4) [0x2b13b88d2974]
> [red0576:07809] [15] openmd_MPI(_ZNSt8ios_base4InitD1Ev+0x49) [0x4744e9]
> [red0576:07809] *** End of error message ***
> _______________________________________________
> Openmd-users mailing list
> Openmd-users at maul.chem.nd.edu
> http://maul.chem.nd.edu/mailman/listinfo/openmd-users
***********************************************
J. Daniel Gezelter
Associate Professor of Chemistry
Director of Graduate Admissions
Department of Chemistry and Biochemistry
251 Nieuwland Science Hall
University of Notre Dame
Notre Dame, IN 46556-5670
phone: +1 (574) 631-7595
fax: +1 (574) 631-6652
e-mail: gezelter at nd.edu
web: http://www.nd.edu/~gezelter
************************************************
More information about the Openmd-users
mailing list