[gmx-users] mdrun with append option

[Justin A. Lemkul]

Sai Pooja wrote:
>  
> On Wed, Feb 2, 2011 at 3:15 PM, Mark Abraham <Mark.Abraham at anu.edu.au 
> <mailto:Mark.Abraham at anu.edu.au>> wrote:
> 
>     On 3/02/2011 6:15 AM, Sai Pooja wrote:
> 
>         The problem is solved with grompp i.e. I use the -t .cpt option.
>         However, now appending does not work. I remember Mark said in a
>         previous mail that a certain environment variable can allow
>         appending to happen even in such cases. I would liek to try that
>         out.
> 
>     No, I said that an environment variable can override the mechanism
>     that blocks ensemble changes in mdrun.
> 
>  
> So how can I use this environment variable.. I might be asking an absurd 
> question since I don't really understand what an environment variable 
> is. But I would definitely liek to experiment with it, since I am in the 
> process of trying out these different options and figuring out which 
> would be the best.
>  

http://en.wikipedia.org/wiki/Environment_variable

I think the pertinent one in this case is GMX_ALLOW_CPT_MISMATCH.  There is a 
reason these aren't well-documented; they probably shouldn't be used in most 
cases.  You should have seen a very specific error message in your .log file or 
screen output indicating that this situation was relevant 
(src/gmxlib/checkpoint.c, around line 1606).

> I also need to understand something. What exactly does the tpbconv do 
> when only -s and -nsteps or -extend options are supplied - it seems that 
> it takes all the information(mass, topology, restraints) from the 
> previous tpr file and just changes the init_step parameter and the 
> number of steps till which the simulation should run.
>  

All it does is modify the number of steps specified in the input file; init_step 
should be untouched.  The step from which the simulation is continued is in the 
.cpt file.

> Now if that is the case, I am still unable to understand that if the cpt 
> file is NOT provided to mdrun (or a mismatched one is provided), how 
> does mdrun obtain the coordinates, velocities, box-dimensions of the 
> last frame. If it doesn't use the ones of the last frame, what does it 
> really use?
>  

These are two cases.  If (1) an invalid .cpt file is provided, the simulation 
should stop with a fatal error (in checkpoint.c, described above); if (2) a 
checkpoint is not provided at all, then a completely new simulation is started, 
and the "last frame" is non-existent.  The simulation begins at time zero.

> If it gets them from the new_tpr file, and the new_tpr file gets it from 
> previous_tpr file via tpbconv, then how does that ensure continuation 
> from the last frame, because the previous_tpr file might have been 
> compiled even before the simulation started. And as far as I know, it is 
> purely an input file to mdrun and has no information on the last 
> coordinates/velocities of the mdrun.

If you provide a .tpr file to mdrun and the checkpoint file is invalid, the 
simulation should have stopped, per the fatal error given in checkpoint.c 
(described above).  The contents of your .log file should make clear exactly 
what mdrun is doing and where it's starting.

In the case you describe, "new.tpr" and "previous.tpr" differ only in the number 
of steps, therefore the state contained therein corresponds to the beginning of 
the simulation, i.e. time zero.

-Justin

>  
> You may have answered this before but I have tried and failed in 
> understanding. I would request you to help me in understanding the 
> above. I would really appreciate it.
>  
> Regards
> Pooja
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================