[gmx-users] g_msd (option -trestart, segmentation fault with -mol)

Vincent Ballenegger vcb25 at cam.ac.uk
Tue May 20 19:45:01 CEST 2003

Dear gmx-users,

I used g_msd to analyse a 400fs run of the spc water model provided in 
/gromacs/tutor/water and I'm wondering about the following 

1) There seems to be a bug when the option -trestart is used together 
with -mol.
Indeed "g_msd -trestart 100" works, but "g_msd -mol -trestart 100" leads 
to a segmentation fault when reading frame 100 corresponding to time 
100ps. This problem was already mentioned in the mailing list in April 
2002, but got apparently no answer.

2) Why is the mean square displacement so nicely proportional to t even 
for small t? Shouldn't it be the case only when t goes to infinity?

3) [Formerly a question, but I found the answer to it. Nothing helps as 
much as explaining what's the problem in full details!]
It took me a long time to understand the effect of the option -trestart. 
Since I'm not the only one
[  for ex. David Bostick asked:
    >Also, what is the -trestart option for? If I were to guess, I would 
say it
    >is the time interval over which the average msd calculation is 
started over
    >again. Is this correct? If this is true then -beginfit and -endfit 
    >be from 0 to the trestart time? I'm kind of confused about what these
    >options mean.
   (no one apparently answered this message posted in Sep. 2002)
I think it would be worth to add a few lines about this option in the 
manual. What confused me in the beginning was the fact that the msd 
curve stills spans the 400ps period when using the option -trestart 200. 
I understand now that instead of stopping the curve at 200ps (the option 
-trestart making this part of the curve to have a statistics twice 
better),  g_msd continues the curve up to 400ps by using data points 
with simple statistics (i.e. the one we would have without the option 
-trestart). It's a logical thing to do, but it can be confusing at first 

Thanks for helping me on questions questions 1 and 2.

Vincent Ballenegger

More information about the gromacs.org_gmx-users mailing list