[gmx-users] Timestep

[Erik Lindahl]

Hi,

In principle Gromacs should never just crash with a segmentation  
fault, but at least give you a (perhaps cryptic) error message and  
exit somewhat gracefully.

As far as I know there is only one exception to this: If you are  
using tabulated interactions the table can only be of finite size,  
and thus there will be a cutoff beyond which there is no data.  
Unfortunately it is not possible to determine this automatically, so  
you have to set the table extension beyond your normal cutoffs in the  
mdp file (table-ext).   The obvious question is of course why we  
don't do a check for the distance before the table lookup, but that  
would be a conditional statement in the very innermost nonbonded  
loop, and drastically reduce performance.

I don't think your system will explode completely with 3fs steps  
though, so the best options might be to compile with debug flag (./ 
configure CFLAGS="-O3 -g") and run the simulation in a debugger to  
see where it crashes.

Cheers,

Erik


On Aug 7, 2007, at 8:30 PM, Xavier Periole wrote

>
> It is not possible to answer your question from the information you  
> give.
> A lot of things can cause a segmentation fault. From compilation to
> your particular system. Put a search on the user-list, you'll get  
> an idea
> of possible problems.
>
> It depends what you will be looking at but I would suggest to stick
> to a 2 fs time step.
>
> XAvier
>
>>  Thanks to Xavier for his prompt reply..
>>  Actually my objective is a bit longer[50-60ns] but the problem is  
>> that
>> even if i cant change the time-step to 0.003 it is exiting saying
>> segmentation fault then how is it possible with 1fs [or 4 fs as
>> suggested] which i want to do to reduce time.
>>  I wanted to know where the problem is... why cant i increase the  
>> time
>> step and the only error shown is 'segmentation fault'..so is the  
>> reason
>> in compilation or coding or machine precision or is it with molecular
>> system wise thats what i was trying to understand..
>> Thanks again..
>> Ramya.
>>>
>>> 10 ns of a solvated protein of regular size should not require much
>>> computing time.
>>>
>>> Increasing the time-step above 2 fs is a solution to increase the
>>> speed of the run but then you face the problem of not integrating
>>> the fast movement correctly and this can end up in large forces
>>> and then you run crashes.
>>> You could use dummy hydrogens (see manual) and heavy water and
>>> use a 4 fs time step. This is ok for a simple protein in a box
>>> of water.
>>>
>>> But again 10ns of a protein solvated is not a long simulation.
>>>
>>> XAvier
>>>>
>>>>    I am trying to perform a simulation run of 10ns for my system
>>>> [protein+waterbox].The system works fine if i use timestep[dt=0.002
>>>> fs].   To balance the run time and time-consumption, when we try a
>>>> .mdp file with slight increase of timestep i.e., to 0.003 fs;  
>>>> grompp
>>>> works fine and generates input tpr, but when going further with  
>>>> mdrun,
>>>> the program crashes, giving error:
>>>>     "Segmentation Fault"
>>>>
>>>> So i have to increase the number of steps[nsteps] to achieve  
>>>> required
>>>> run
>>>> time.But this'll be time consuming.I am unable to figure out any
>>>> alternative.
>>>> I am using Gromacs 3.3 on IBM Power5 machine with SLES9.
>>>> Please suggest me the required to be done.Thanks in advance.
>>>>
>>>> The .mdp file i've used:
>>>>
>>>> integrator               = md
>>>> ;Total simulation time: 5000 ps
>>>> :time step in femtoseconds
>>>> dt                       = 0.002
>>>> ;number of steps
>>>> nsteps                   = 2500000
>>>> nstxout                  = 100
>>>> nstvout                  = 100
>>>> nstlog                   = 10
>>>> nstenergy                = 10
>>>> nstxtcout                = 10
>>>> ;xtc_grps                = prot
>>>> ;group(s) to write to energy file
>>>> energygrps               = protein non-protein
>>>> Frequency to update the neighbor list (and the long-range forces,
>>>> ;when using twin-range cut-off's).
>>>> nstlist                  = 10
>>>> cut-off distance for the short-range neighbor list
>>>> rlist                    = 0.8
>>>> ;treatment of electrostatic interactions
>>>> coulombtype              = PME
>>>> rcoulomb                 = 0.9
>>>>
>>>>
>>>>
>>>> *********$$$*********
>>>>
>>>> Ramya Cherukupalli,
>>>> Final year,
>>>> M.S Bioinformatics,
>>>> IIIT - H.
>>>>
>>>>
>>>> --
>>>> This message has been scanned for viruses and
>>>> dangerous content by MailScanner, and is
>>>> believed to be clean.
>>>>
>>>> _______________________________________________
>>>> gmx-users mailing list    gmx-users at gromacs.org
>>>> http://www.gromacs.org/mailman/listinfo/gmx-users
>>>> Please search the archive at http://www.gromacs.org/search before
>>>> posting!
>>>> Please don't post (un)subscribe requests to the list. Use the
>>>> www interface or send it to gmx-users-request at gromacs.org.
>>>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>>>
>>> -----------------------------------------------------
>>> XAvier Periole - PhD
>>>
>>> 1- Institute of Molecular Assemblies
>>>       City University of New York - USA
>>> 2- Molecular Dynamics-Group
>>>       University of Groningen - The Netherlands
>>> http://md.chem.rug.nl/~periole
>>> -----------------------------------------------------
>>> _______________________________________________
>>> gmx-users mailing list    gmx-users at gromacs.org
>>> http://www.gromacs.org/mailman/listinfo/gmx-users
>>> Please search the archive at http://www.gromacs.org/search before  
>>> posting!
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-users-request at gromacs.org.
>>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>>>
>>> --
>>> This message has been scanned for viruses and
>>> dangerous content by MailScanner, and is
>>> believed to be clean.
>>>
>>>
>> *********$$$*********
>> Ramya Cherukupalli,
>> Final year,
>> M.S Bioinformatics,
>> IIIT - H.
>> -- 
>> This message has been scanned for viruses and
>> dangerous content by MailScanner, and is
>> believed to be clean.
>> _______________________________________________
>> gmx-users mailing list    gmx-users at gromacs.org
>> http://www.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive at http://www.gromacs.org/search before  
>> posting!
>> Please don't post (un)subscribe requests to the list. Use the www  
>> interface or send it to gmx-users-request at gromacs.org.
>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
> -----------------------------------------------------
> XAvier Periole - PhD
>
> 1- Institute of Molecular Assemblies
>      City University of New York - USA
> 2- Molecular Dynamics-Group
>      University of Groningen - The Netherlands
> http://md.chem.rug.nl/~periole
> -----------------------------------------------------
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before  
> posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php