From: Joshua Adelman (jadelman_at_berkeley.edu)
Date: Mon Aug 03 2009 - 10:38:35 CDT
If you have a set maximum time you can run on your calculation on your
cluster for, you should be able to estimate how many steps you can
generate in that time and then just submit jobs of that length or
less, then restart the next run from the previous run. I'm not quite
sure what the <max_time> and <quantum_time> parameters are in your
example, but you can just set the number of steps in the namd config
file. You could certainly write a script to automatically generate
config files and submit them at some time interval while checking to
see if the previous simulation has ended.
Maybe I'm missing something, but otherwise this seems like a pretty
straightforward problem that doesn't require a fancy solution.
On Aug 3, 2009, at 8:30 AM, Mikhail Suyetin wrote:
> Dear Josh, everybody,
> I need to perform a long simulation, but I cannot use my cluster all
> the time. Therefore, If I want to make a simulation for a day I'll
> have to split time of calculation. For example:
> ./run_mvapich.sh <number_of_processors> my_folder/1.namd <max_time>
> Sincerely yours
> Mikhail Suyetin
> 2009/8/3 Joshua Adelman <jadelman_at_berkeley.edu>:
>> I'm not sure exactly what you are asking. If you want to take a long
>> simulation you have run and divide the resulting trajectory into
>> pieces, you should use catdcd and use the '-first' and '-last'
>> flags. On the
>> other hand, if you want to run a simulation for some length of
>> time, L, but
>> want to run it in N pieces, then you would just run a number of
>> of length L/N and restart the n+1 simulation with the coordinates and
>> velocities from simulation n. Or do you want to do something else?
>> If so,
>> you'll have to elaborate.
>> Hope that helps.
>> On Aug 3, 2009, at 6:26 AM, Mikhail Suyetin wrote:
>>> Hi everybody!
>>> Does anybody have a script for performing calculations divided by
>>> quanta, i.e. I would like to divide total time of MD simulation into
>>> several parts.
>>> Sincerely yours,
>>> Mikhail Suyetin
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:06 CST