Hi,
On Wed, Jan 19, 2011 at 10:38:52PM -0500, Jason Swails wrote:
> On Wed, Jan 19, 2011 at 9:46 PM, Tyler Luchko <tluchko.rci.rutgers.edu>wrote:
> > On 2011-01-19, at 6:12 PM, Jason Swails wrote:
> >
> > > rism3d/ala && ./Run.ala.hnc && ./Run.ala.pse1 -- exits in error with
> > sander
> > > printing the message [ ERROR> solvcut must > 0. ]
> >
> > I have fixed this but haven't pushed it out yet. I'll do this ASAP.
Thanks this one is fixed.
> > > Amber (parallel, 2 threads):
> > >
> > > All sander.RISM.MPI tests failed with the message [ Fatal error in
> > > MPI_Comm_rank: Invalid communicator, error stack: ]
> >
> > What are your configure options? I'm not seeing this with GNU or Intel
> > compilers.
I also see failures with all these tests using gnu 4.1.2 and mvapich2-1.5:
./Run.ala
SANDER: 3D-RISM
mpiexec: Warning: task 0 died with signal 11 (Segmentation fault).
mpiexec: Warning: tasks 1-3 died with signal 15 (Terminated).
This is more circumstantial evidence against mpich2.
scott
> Ubuntu Linux x86_64. Intel compilers 11.1.069 with Intel MKL (same
> version). mpich2 version 1.2.1b.
>
> mpicc -show = icc -fpic -I/mpi/intel-11.1.069/mpich2-1.2.1/include
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib -lmpich -lopa -lpthread -lrt
>
> mpif90 -show = ifort -I/mpi/intel-11.1.069/mpich2-1.2.1/include
> -I/mpi/intel-11.1.069/mpich2-1.2.1/include
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib -lmpichf90 -lmpichf90 -lmpich -lopa
> -lpthread -lrt
>
> mpif77 -show = ifort -fpic -I/mpi/intel-11.1.069/mpich2-1.2.1/include
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib
> -L/mpi/intel-11.1.069/mpich2-1.2.1/lib -lmpich -lopa -lpthread -lrt
>
> I'm not sure why mpif90 doesn't have the flag for PIC... I can change mpif90
> to mpif77 and see if that makes a difference.
>
> Just as a note, this error does not crop up on my Mac OS X, GNU 4.4.5,
> OpenMPI 1.4.1 setup. I can check my Mac to see if it's an mpich2 issue. I
> remember just prior to the Amber 11 release there was a similar issue that
> surfaced on the developer list that Mark Williamson found and fixed
> pertaining to MPICH2. Perhaps it's that?
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sun Jan 23 2011 - 17:00:02 PST
