Just some additional info.
I commented out the single test Run.cnstph from the Makefile, and
re-ran the test suite.
Both test.parallel.MM and test.parallel.QMMM ran without significant
errors. I did notice that several tests required that they be run with
only 2 processes, and therefore were skipped since I ran with 4.
Cheers,
--Jim Parker
UTSA Physics
Jim Parker wrote:
> I completed the install/testing of Amber 10 on my Ubuntu system.
> Thanks Ross...
>
> I installed the parallel version with no errors with only a single
> manual modification of the config_amber.h file
>
> using the default MPI (openmpi) for Ubuntu
> >apt-get install libopenmpi-dev libopenmpi1 openmpi-bin openmpi-common
> openmpi-doc
>
> which puts all MPI includes and libraries in /usr/lib/openmpi.
> so defining MPI_HOME appropriately
> >export MPI_HOME=/usr/lib/openmpi
>
> I configure the build with
> > ./configure_amber -openmpi gfortran
>
> which I then edit, config_amber.h manually because the wrapper
> compilers mpicc and mpif90 are installed into /usr/bin and not
> $MPI_HOME/bin as the ./configure_amber script expects
>
> in config_amber.h:
>
> change FC to
> FC=/usr/bin/mpif90
>
> and change LOAD
> LOAD= /usr/bin/mpif90 $(LOCALFLAGS) $(AMBERBUILDFLAGS)
>
> then compile with
> >make parallel
>
> The build completed without errors.
>
> So moving to the test
> >cd $AMBERHOME/test
> >export DO_PARALLEL="mpirun -np 4"
> >make test.parallel.MM
>
> runs through several successful tests, but eventually fails at
> cd cnstph && ./Run.cnstph
> [ubuntu1:28131] *** Process received signal ***
> [ubuntu1:28133] *** Process received signal ***
> [ubuntu1:28133] Signal: Segmentation fault (11)
> [ubuntu1:28133] Signal code: Address not mapped (1)
> [ubuntu1:28133] Failing at address: (nil)
> [ubuntu1:28133] [ 0] [0x4001a410]
> [ubuntu1:28132] *** Process received signal ***
> [ubuntu1:28132] Signal: Segmentation fault (11)
> [ubuntu1:28132] Signal code: Address not mapped (1)
> [ubuntu1:28132] Failing at address: 0x6a7ef9db
> [ubuntu1:28131] Signal: Segmentation fault (11)
> [ubuntu1:28131] Signal code: Address not mapped (1)
> [ubuntu1:28131] Failing at address: 0x5
> [ubuntu1:28132] [ 0] [0x4001a410]
> [ubuntu1:28132] [ 1]
> /home/surfer/tmp/amber10/exe/sander.MPI(sander_+0x6f45) [0x80f49fd]
> [ubuntu1:28132] [ 2]
> /home/surfer/tmp/amber10/exe/sander.MPI(MAIN__+0x106c) [0x80eda34]
> [ubuntu1:28133] [ 1]
> /home/surfer/tmp/amber10/exe/sander.MPI(sander_+0x6f45) [0x80f49fd]
> [ubuntu1:28132] [ 3]
> /home/surfer/tmp/amber10/exe/sander.MPI(main+0x39) [0x841d3f9]
> [ubuntu1:28133] [ 2]
> /home/surfer/tmp/amber10/exe/sander.MPI(MAIN__+0x106c) [0x80eda34]
> [ubuntu1:28132] [ 4]
> /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe5) [0x402ce685]
> [ubuntu1:28133] [ 3]
> /home/surfer/tmp/amber10/exe/sander.MPI(main+0x39) [0x841d3f9]
> [ubuntu1:28132] [ 5] /home/surfer/tmp/amber10/exe/sander.MPI [0x807aaa1]
> [ubuntu1:28133] [ 4]
> /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe5) [0x402ce685]
> [ubuntu1:28132] *** End of error message ***
>
> Any help on how to proceed would be appreciated.
> searches of the archives for Run.cnstph returned errors that were
> different from this.
>
> Cheers,
> --Jim
> UTSA Physics
>
>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Apr 12 2009 - 01:17:16 PDT
