AMBER Archive (2006)
Subject: RE: AMBER: Problem of QM/MM calculation with amber 9 parallel version
From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Wed Nov 01 2006 - 15:36:17 CST
With regards to your first question the QM/MM code is not completely
parallelized. Although I'm not sure what you mean by just running with one
process. - where did you get the information from that it is running 1
If I explain how QM/MM currently works in parallel this might make it a bit
clearer. At present during a QM/MM simulation the whole of the MM part of
the simulation is parallel and all parts of the QM simulation are parallel
with the exception of the matrix diagonalization (see Dave I am learning ;-
)) which is done only on the master processor. For systems that have a
large MM region, a fairly small QM region (50 - 60 atoms) and use a fairly
complex QM potential (I.e. they have pme or GB turned on for QM) then you
can get some scaling to about 4 to 8 processors or so at which point the
serial nature of the diagonalization prevents any more scaling. This would
be say for 30,000 MM atoms and 50 QM atoms. If you have many more QM atoms
(>80 or so) then the matrix diagonalization dominates
so you won't see much benefit on multiple cpus.
Hence the benefits of running QM/MM in parallel really depends on the type
of simulation you are running. Typically I find on a 4 processor SMP
machine you can run reasonably in parallel for a typical QM/MM simulation
in explicit solvent.
If however, you find that it does not run in parallel at all (no atom
division data in the output file etc) then something is most likely wrong.
With regards to the peptide correction this is based on an approach that
was originally adopted in MOPAC and I believe that it is generally accepted
that you need this correction for AM1 or PM3 when dealing with peptide
linkages. Here is a discussion of it in the mopac manual:
there are also some cautions here
I am not aware if there is a paper discussing this. You may want to ask the
question on CCL to see whether MM corrections to peptide barrier heights in
semi-empirical calculations are discussed in any publications.
All the best
| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.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
Sent: Wednesday, November 01, 2006 11:50
Subject: AMBER: Problem of QM/MM calculation with amber 9 parallel version
Dear Amber users,
I have two question about the QM/MM calculations.
First, I installed Amber 9 to a parallel version.
It's working with multi-process in simulating classical MD.
but in simulating with multi-process in QM/MM calculation, sander are not
just runing with one process.
Is it normal?
Next, I want to know about PEPTIDE_CORR option in QM/MM calculation.
I have calculated the system that consist of small QM area(peptide) and
remaining MM area.
After some time in simulation, backbone angle(omega angle) was changed to
Usually, that peptide are not exist in cis form.
To maintain the peptide angle to trans form, I need to use PEPTIDE_CORR
I want to know why this phenomenon happen and know papers about this
"우리 인터넷, Daum" http://www.daum.net 『평생쓰는 무료 한메일넷』
-- The AMBER Mail Reflector To post, send mail to amber_at_scripps.edu To
unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu