Re: failure in RATTLE

From: Axel Kohlmeyer (akohlmey_at_cmm.chem.upenn.edu)
Date: Wed Dec 17 2008 - 09:52:13 CST

On Wed, 17 Dec 2008, Sangamesh B wrote:

SB> One of our customer gave this input file for benchmark, saying that
SB> job has run for 7 days with 16 processes on his cluster.

yes. but was it the _exact_ same input? i doubt it.
be it as it may, the output indicates that it is very
likely that the input is starting from a high potential
energy configuration and that this does not run stable.

[...]

SB> Ok. But what might be the wrong with the current install? I've used

depending on the optimization level of compilers codes can
give slightly different results. for a marginal input that
may just be enough. as i was writing before, somebody in your
line of work should be well aware of that.

[...]

SB> Initial lines indicate that it's using mpirun:
SB>
SB> $ cat npt_result_out_ll
SB>
SB> Running on 4 processors: /opt/apps/namd26_gnu/Linux-amd64-MPI/namd2 npt02.inp
SB> charmrun> mpirun -np 4 /opt/apps/namd26_gnu/Linux-amd64-MPI/namd2 npt02.inp
SB> Charm++> Running on MPI version: 2.0 multi-thread support: 0/0
SB> Charm++> Running on MPI version: 2.0 multi-thread support: 0/0
SB> Charm++> Running on MPI version: 2.0 multi-thread support: 0/0

this is not correct. how else would you get each output 4 times?
this only indicates, that you _compiled_ for MPI.

if you look a little bit deeper in the output file (from the mail
that didn't make it through the mailing list), you'll see that
NAMD reports running on one processor four times.

cheers,
   axel.

-- 
=======================================================================
Axel Kohlmeyer   akohlmey_at_cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:50:16 CST