[AMBER-Developers] clean stop for TI bug?

From: G. Andres Cisneros <andres.chem.wayne.edu>
Date: Tue, 16 Apr 2013 11:03:17 -0400

Hi Dave,

I am teaching a grad computational chemistry class this semester and one
of the students is doing TI with AMBER.

He ran into a problem with his input because he didn't follow the
protocol and created prmtops with different numbers of atoms. However,
when he ran the calculation he only got an MPI crash report:

-------
mpirun -np 2 sander.MPI -ng 2 -groupfile model_step1.group

 Running multisander version of sander Amber12
    Total processors = 2
    Number of groups = 2
 
Fatal error in PMPI_Bcast: Message truncated, error stack:
PMPI_Bcast(1478)......................: MPI_Bcast(buf=0x97d96f8,
count=18, MPI_DOUBLE_PRECISION, root=0, comm=0x84000003) failed
MPIR_Bcast_impl(1321).................:
MPIR_Bcast_intra(1119)................:
MPIR_Bcast_scatter_ring_allgather(962):
MPIR_Bcast_binomial(145)..............:
MPIDI_CH3U_Receive_data_found(129)....: Message from rank 0 and tag 2
truncated; 152 bytes received but buffer size is 144
--------

I am not sure if this is extremely important or if I should send it to
the amber-devel list, but maybe it would make sense to add a check when
ifce=1 and ng > 1 to make sure the prmtops have the same number of
atoms, and if not to make a clean stop?

Best regards,

Andrés

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Apr 18 2013 - 12:30:02 PDT
Custom Search