AMBER Archive (2005)

Subject: RE: AMBER: serial and parallel

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Sun Apr 24 2005 - 13:49:38 CDT


Hi Aknb
 
To answer your first question it is perfectly fine to restrt a serial run in
parallel this will cause no problems whatsoever. Just be sure that you
parallel installation passes the AMBER tests to ensure it is functioning ok.
You can do this as follows:
 
Assuming you are using lam, adjust for your parallel setup if necessary:
 
lamboot list_of _hosts
cd $AMBERHOME/test
export DO_PARALLEL="mpirun -np 4" (Or however many cpus you want etc...)
make
 
Any potential failures should be just simple round off error. Or specific
tests that have messages saying they can't be run in parallel.
 
Just one thing you should be aware of. If you run a simulation in serial and
then the same simulation in parallel you won't necessarily get the same
trajectory. Initially they will track but after a while the trajectories
will diverge due to roundoff error / orders of execution etc. This doesn't
mean any one is "less" right than the other, just that it is exploring a
different region of phase space. There is much discussion on this on the
amber mailing list archive.
 
All the best
Ross
 
/\
\/
|\oss Walker

| Department of Molecular Biology TPC15 |
| The Scripps Research Institute |
| Tel:- +1 858 784 8889 |
| EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk/ | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

  _____

From: owner-amber_at_scripps.edu [mailto:owner-amber_at_scripps.edu] On Behalf Of
Aknb
Sent: 24 April 2005 04:36
To: amber_at_scripps.edu
Subject: AMBER: serial and parallel

Dear Amber Users,
 
i have two questions to ask and really hope to get some feedbacks from all
of you
 
1) is it okay to combine serial and parallel runs together? i mean, for
example, i did 20 ns of simulation in serial and from 21-40 ns was later
done in parallel for the same system. it's not consistent, but will that
leads to faulty results?
 
2) my second question is regarding the MM-PB/GBSA calculation, is there any
hard rules for the input to be trajectories collected from explicit solvent
simulation? can i use trajectories collected from in vacuo simulation as an
input for MM-PB/GBSA?
 
thank you very much in advance
 
 
 

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu