[gmx-users] question about continuation using tpbconv and mdrun -cpi

[Baofu Qiao]

Hi Mark,

Thanks!
Because the maximum time for one single job is set to be 24hours on the
cluster I'm using, I want to make sure which is the best way to continue
the gmx jobs. I wonder how strong effect "mdrun -cpi" has?  From the
introduction of mdrun, it seems that there are some EXTRA energy frames,
but for the trajectory file (.xtc), there is no extra frames? Am I right?

"mdrun -h
--> The contents will be binary identical (unless you use dynamic load
balancing), but for technical reasons there might be some extra energy
frames when using checkpointing (necessary for restarts without appending)."

best wishes,

Mark Abraham wrote:
> Baofu Qiao wrote:
>> Hi all,
>>
>> Recently I made some tests on the continuation of gmx 4.0.5. Firstly, I
>> run a complete simulation as the reference. In the full.mdp file,
>> gen_seed=123456 & continuation=yes are used to build the same initial
>> structure.  In all the simulations, "mdrun -dlb no" is used because I
>> want to reproduce exactly the same potential energy.
>>
>> 1. Using tpbconv.  (part1.mdp is the same as full.mdp except the
>> "nsteps")
>>   1a) grompp .... -s part1
>>       mpirun -np 32 mdrun -deffnm part1 -dlb no
>>       tpbconv -s part1.tpr -f part1.trr -e part1.edr -extend 100 -o
>> part2
>>       mpirun -np 32 mdrun -deffnm part2 -dlb no
>>       eneconv -f part1.edr part2.edr -o part_all.edr
>>   1b) tpbconv -s part1.tpr -extend 100 -o part2.tpr
>>       mpirun -np 32 mdrun -deffnm part2 -cpi part1.cpt -dlb no
>>       eneconv -f part1.edr part2.edr -o part_all.edr
>> The potential energy is generated from g_energy, and compared. In this
>> method, I met two problems:      Q.a) The potential energies of
>> part2  in both 1a and 1b are not
>> exactly the same as the reference potential!  And also the potential of
>> part2 in 1a is different from that in 1b. (Both potentials of part1 from
>> 1a and 1b are identical to the reference.)
>>      Q.b) The RMSD of the potential energies of part_all.edr is very big
>> (1 order of magnitude bigger than the corresponding one from the
>> separate .edr files)
>>
>> 2) using mdrun -cpi -append
>>         grompp -s full2
>>         mpirun -np 32 mdrun -deffnm full2 -dlb no  (stop the job at some
>> time point, then run the following)
>>         mpirun -np 32 mdrun -deffnm full2 -dlb no  -cpi full2_prev.cpt
>> -append
>> The second (-append) section of the potential energy is also different
>> from the result of the reference potential, even though I am using the
>> same full.mdp file.  (the potential in the former section is identical
>> to the reference)
>>
>> Then, how to reproduce exactly the potential?
>
> While "mdrun -dlb no" is a good start, "mdrun -reprod" takes care of
> some other things too...
>
>>  It is said that the
>> continuation from "mdrun -cpi" is binary identical. 
>
> Not quite... it's an exact restart from the old point in the ensemble,
> but the manner in which mdrun does future integration is only
> reproducible if you force it to be so.
>
>> However, it seems
>> not in my tests. What's the problem of the very big RMSD of the
>> potential from eneconv?
>
> Not sure.
>
> Mark
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.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
>