Running NAMD simulations

NAMD requires certain input files that specify the system coordinates and molecular structure and a configuration file that lists simulation options. See Sections 3 and 4 for more details concerning these files.

In a distributed simulation with full electrostatics using DPMTA, there will be three processes running per node:

• namd -- the NAMD process
• dpmta_slave -- the DPMTA slave process
• pvmd -- the PVM daemon process

Suppose we wish to start a distributed NAMD simulation on, say, the four hosts foo1, foo2, foo3, and foo4. From host foo1 we could invoke the PVM daemon and add the other hosts, execute NAMD using configuration file my-config-file, and halt PVM after NAMD exits. The sequence of commands looks like:

foo1> pvm
pvm> add foo2 foo3 foo4
3 successful

...  more output from PVM appears here  ...

pvm> quit

pvmd still running.
foo1> namd my-config-file

...  a whole lot of output from NAMD appears here  ...

foo1> pvm
pvmd already running.
pvm> halt
foo1>

Now, suppose we wish to start a multi-processor NAMD simulation on, say, the host foo. From host foo we invoke the PVM daemon, execute NAMD using configuration file my-config-file along with the number of nodes to use, and halt PVM after NAMD exits. (Note that NAMD needs to have been compiled with the -DSPECIFY_NUM_NODES option on the NAMD_CFLAGS_ARCH line in Make.arch.) The sequence of commands looks like:

foo> pvm
pvm> quit

pvmd still running.
foo> namd my-config-file 4

...  a whole lot of output from NAMD appears here  ...

foo> pvm
pvmd already running.
pvm> halt
foo>

It might also be necessary to set the environmental variable PVMBUFSIZE that indicates the amount of space to devote to message buffering. We suggest anywhere from 4 to 32 MB depending on the size of the molecular system. See the PVM documentation for more details.