Re: [AMBER] CPPTRAJ: Is it a program that requires a large memory?

From: Jason Swails <>
Date: Mon, 27 Aug 2012 06:59:52 -0400

On Mon, Aug 27, 2012 at 3:06 AM, Catein Catherine <>wrote:

> Dear Experts,
> I am using cpptraj to analysis my mdcrd files (a total of 6 G gzipped
> mdcrd files).
> However, the jobs stopped in the middle of the analysis without error
> message, e.g.. The jobs just stopped like this:

----- [md0.mdcrd.gz] (1-100, 1) ----- 0% 10% 20%
> I am using the supercomputer to do the analysis, the administrator said it
> is because of the large memory required by the cpptraj.

Some actions in cpptraj require large amounts of memory (e.g.
dihedralcluster and 2drms, I think), since they have to store all
coordinates that are read in. I think hbond also requires a large amount
of memory.

For the most part, though, the memory usage is fairly modest, since all
that has to be stored at any given point is just a single set of
coordinates, all of the data in the topology file, and all of the data you
plan to write out at the end of the run.

However, it depends on what you mean by 'large memory' -- for a big
topology file, you have a lot of data that needs to be stored from the HPC,
and each frame is fairly large.

> Could you mind to share your experience with me how to solve this problem?

Request more memory, or analyze fewer frames at a time. Or, in the case of
hbond analysis, use a smaller list of acceptors and donors.

> My cpptraj file is as
> follows:===================================================================================parm
> all.prmtoptrajin md0.mdcrd.gz........trajin md35.mdcrd.gzautoimage anchor
> :37-460trajout qsub.netcdf netcdf .C1',CArms first out all_all.rms
> .C1',CAatomicfluct out all.bfactor byres bfactoratomicfluct out
> all.nobfactor .C1',CA byresaverage all_average_nobox.pdb pdb noboxmatrix
> correl .C1',CA out all.corr byreshbond out all_hbond.dat avgout
> all_hbond.datcluster C1 clusters 3 epsilon 4.0 out all.cnumvtime.dat
> summary all.avg.summary.dat clusterout all.clustertraj clusterfmt
> netcdf===================================================================================

The lack of carriage returns makes this input file too difficult to parse.
 Please try and make sure that carriage returns show up in your emails to
make it easier for us to help.


Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER mailing list
Received on Mon Aug 27 2012 - 04:30:03 PDT
Custom Search