Re: CCL:Reusing the NAMD code to build C++ simulation program
- From: masakatsu-ito^at^aist.go.jp
- Subject: Re: CCL:Reusing the NAMD code to build C++ simulation
program
- Date: Wed, 13 Aug 2003 18:25:03 +0900 (JST)
Dear Dr. Phillips,
Thank you so much for your helpful advice. Now I am planning to
rewrite a part of our C++ framework in Charm++ to replace the
mainmodule (and main() ?) of the NAMD code with our code.
To build a replica exchange molecular dynamics (REMD) or REM
simulation, I have to instantiate a set of molecular modeling objects
which return a potential energy and force vector as a function of atom
positions. I, however, am afraid that a Tcl interpreter does not
provide a way for creating a set of namd2 objects to which our
framework delegates force field calculations. Thus I am trying to
understand which objects in the NAMD code performs the force field
calculation and how to instantiate those objects.
There are two points in the NAMD code which puzzle me. First, the
main() function does not seem to be invoked in the Charm++ execution
model, because the Charm++ programming manual says that the Charm
Kernel starts the execution by creating a mainchare instance. The
manual does not show any example code including a main() function so
that I cannot see how a main() behaves in Charm++.
The second is the calling logic for force field computation. I think
that the Node object is responsible for all the actual calculation as
if the functionalities of a main() function for each CPU are divided
into its member functions. And the Sequencer object instantiated in
the Node seems to integrate the equation of motion to update the atom
positions and velocity vectors. Thus, in our code, I would like to
set the atom positions at the Patch objects in the Sequencer, to
trigger the force field computation, and to get the resultant force
vectors from the Patch objects. But I cannot see which member
function triggers the force field computation. Could you tell me which
member function should I look into?
It is exciting for me to look into the NAMD code and to see how its
decomposition scheme and load balancing functionality are implemented.
I would be very grateful for any hint or pointer to the documents for
the NAMD implementation.
Sincerely,
Masa
Masakatsu Ito , Ph.D
Grid Technology Research Center
National Institute of Advanced Industrial Science and Technology
Tsukuba Central 2, Tsukuba, Ibaraki 305-8568 Japan
Phone : +81-29-861-5730 Fax : +81-29-861-5301
E-mail masakatsu-ito^at^aist.go.jp
URL http://met.sourceforge.net/
From: Jim Phillips <jim^at^ks.uiuc.edu>
Subject: Re: CCL:Reusing the NAMD code to build C++ simulation program
Date: Mon, 11 Aug 2003 16:55:18 -0500 (CDT)
Message-ID:
<Pine.GSO.4.40.0308111642200.8311-100000^at^verdun.ks.uiuc.edu>
> Hi,
>
> We are looking at making NAMD more modular, but it isn't there yet. As
> you observe, Charm++ runs through the entire control structure. You might
> be able to graft the entire namd2 program into your code as a single large
> simulation object by replacing main() in mainfunc.C with your interface.
> NAMD has a Tcl interpreter running at the top level that could maybe be
> used to accept commands from your framework. Otherwise, it should be
> possible to implement REM with a small amount of network code in a NAMD
> Tcl script (and possibly some new Tcl commands). Good luck.
>
> -Jim
>
>
> On Sat, 9 Aug 2003 masakatsu-ito^at^aist.go.jp wrote:
>
> > Dear Markus,
> >
> > Thank you for your quick reply. As you wrote, Mindy,
> > a sequential version of NAMD, helped me a lot.
> > It was easy to reuse and modify its code.
;
> > Thus I rephrase my question. Could somebody give me an explanation
> > if it is possible to use the NAMD code instead of the Mindy one
> > and how to incorporate its Charm++ code into a C++ framework.
> >
> > From: Markus Dittrich <markus^at^ks.uiuc.edu>
> > Subject: Re: CCL:Reusing the NAMD code to build C++ simulation program
> >
> > > Dear Masakatsu,
> > >
> > > The TCBG group does offer an MD code called
> > > MINDY which is derived form NAMD source code
> > > written solely in C++. Maybe this will be of
> > > help to you.
> > >
> > > http://www.ks.uiuc.edu/Development/MDTools/mindy/
> > > cheers,
> > > MARKUS
> > >
> > > On Fri, 8 Aug 2003 masakatsu-ito^at^aist.go.jp wrote:
> > >
> > > > Dear CCL'ers,
> > > >
> > > > I am wondering if it is easy to reuse the NAMD code to build
C++
> > > > simulation program. My question might sound strange because
NAMD is so
> > > > famous for its high extensibility and object-oriented design
with
> > > > C++. But NAMD is actually coded in Charm++, an extended C++
for
> > > > parallelization, so that the problem is that a programmer
has to deal
> > > > with Charm++ syntax in the NAMD code and its execution model
for
> > > > parallel computation. So, could somebody give me a brief
explanation
> > > > how the NAMD code can be incorporated into C++ framework, or
if it is
> > > > possible.