Why is the timestep incorrect in DCD files?
I assume you are using X-PLOR to read the trajectories. A couple of years
ago, after X-PLOR development stopped, the NAMD DCD format was adjusted
slightly to conform to the CHARMM specification. The difference between
the two is that while X-PLOR used a double for the timestep, CHARMM uses
the same space for a float followed by an int. That int is important
because it indicates if the file contains data about the periodic cell, a
recent addition. It was previously ignored so CHARMM and X-PLOR could
exchange DCD files with only a bad timestep, but now CHARMM rejects X-PLOR
written files unless the timestep is zero.
I would suggest trying to cope with the incorrect timestep. If you can't
the NAMD source distribution contains dumpdcd.c and loaddcd.c, which can
be used to set the headers to whatever you want, and fixdcd.c, which can
be modified to fix your problem (it currently changes X-PLOR to CHARMM).
Why does NAMD 2.1 complain about missing vdw parameters?
FATAL ERROR: DIDN'T FIND vdW PARAMETER FOR ATOM TYPE 0.000000E+00
If you don't specify a segment name, the second column in the psf
file will be blank and NAMD will misread the charge as the atom type.
Fix this by adding name="MAIN" to your segment statement in X-PLOR or
giving the segname as the first argument to GENErate in CHARMM.
Why does NAMD 2.2 complain about CHARMM format?
FATAL ERROR: Structure (psf) file is in CHARMM format; XPLOR format required.
CHARMM can create psf files in either CHARMM format (which uses numbers for
atom types) or X-PLOR format (which uses names for atom types). NAMD uses
X-PLOR format both for historical reasons, and because it is more robust.
If NAMD sees a numerical atom type it tries to be helpful with this message.
The solution is to use "write psf card xplor" when creating the PSF file.
If you did this (or used X-PLOR in the first place) then you probably have
a blank segment name. If you don't specify one, the second column in the psf
file will be blank and NAMD will misread the charge as the atom type.
Fix this by adding name="MAIN" to your segment statement in X-PLOR or
giving the segname as the first argument to GENErate in CHARMM.
Is there an SGI version of NAMD?
There are two, Origin and Origin-MPI. These may work on other SGI's which
have R10000 or later CPUs. The Origin version will run within a single
machine and the MPI version will also run between boxes, assuming SGI's
implementation of MPI supports this.
Why are there both Origin and Origin-MPI versions?
The Origin version uses a shared memory arena to exchange data between
threads. It runs without any special support and scales fine for small
numbers of processors. If you want to run on more that 8 processors,
however, you should get the Origin-MPI version which scales better.
We don't know why the MPI version is faster.
|
