LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] LAMMPS: Inconsistent Results When Using Same Random Seed"
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] LAMMPS: Inconsistent Results When Using Same Random Seed"


From: Axel Kohlmeyer <akohlmey@...24...>
Date: Sat, 12 Aug 2017 21:47:59 -0400



On Sat, Aug 12, 2017 at 8:53 PM, MASATO KOIZUMI <mkoizumi@...5197...09...> wrote:
Dear Dr. Axel Kohlmeyer,

Hello Dr. Kohlmeyer. Thank you for the guidelines.

​you don't seem to be following them very closely, though.​


I am using the LAMMPS version February 10, 2015. The LAMMPS input script that I am running equilibrates the randomly created polymer chains. The data file that this input script reads contains 250000 atoms.

​i don't care about looking into how an over 2.5 year old LAMMPS version is behaving.
i also don't care about having to dig through a word document for not good reason.
 

I looked at the previous mailing list in regards to inconsistent MD trajectories. I am using the same number of processors for every run. Therefore, inconsistent position and velocity coordinates of the atoms should not occur due to the number of processors.

​the *will* occur. after some time. your simulation protocol is making it *extremely* likely. 
you are doing floating point math ...and it is quite possible to have "inexact" operations where random bits (e.g. through thermal noise) enter into the calculation especially when your executable was compiled with any form of compiler optimization.

using DPD enhances this. as soon as there is a difference of one atom being placed into a different subdomain, your trajectories will diverge quickly due to the RNGs being called differently frequent on different subdomains. the more processors you use, the more likely this gets.
 

My question that illustrates my difficulties are described in the attached Word document. For clarity, I also included my LAMMPS input script in the attached Word document. If you require the data file, I would greatly appreciate if you could notify me.

​word documents are very *in*convenient.​ altogether, i don't see any convincing reason to investigate this. your expectation to have identical trajectories is not justified for this kind of simulation protocol and conditions.

axel.

 

Thank you.

Sincerely,

Masato Koizumi

On Sat, Aug 12, 2017 at 2:59 PM, Axel Kohlmeyer <akohlmey@...24...> wrote:
i should add the following.

in order for you to have more productive conversations via this mailing list, please have a look at this http://lammps.sandia.gov/guidelines.html 
several of those apply here. i particularly recommend to search the mailing list archive. the topic of diverging trajectories has been discussed numerous times on 
this mailing list already.

On Sat, Aug 12, 2017 at 4:02 PM, Axel Kohlmeyer <akohlmey@...24...> wrote:


On Sat, Aug 12, 2017 at 3:40 PM, MASATO KOIZUMI <mkoizumi@...3709...> wrote:
Dear Dr. Axel Kohlmeyer,

Hello Dr. Kohlmeyer. Thank you for your e-mail. 

I was curious if you could inform me a solution to obtain the exactly the same results in regards to trajectories if I rerun the same LAMMPS input script. 

I was curious if using the exactly the same random seed will reproduce the same results such as not changing the random seed of the velocity and dpd pairwise potential for every rerun. Thank you.

​same question, same answer. 
your request is like a person going to the doctor complaining about pain, but not saying where exactly it hurts or how much, but expecting relief. 
​so the best i can do is to recommend: take 2 aspirin and only come back when your can tell us more.

​axel.​

Sincerely,

Masato Koizumi

On Sat, Aug 12, 2017 at 11:49 AM, Axel Kohlmeyer <akohlmey@...24...> wrote:


On Fri, Aug 11, 2017 at 7:43 PM, MASATO KOIZUMI <mkoizumi@...3709...> wrote:
Dear All,

Hello. This is Masato Koizumi, and I am using LAMMPS for my MD simulations. 

I am currently running a LAMMPS input script that equilibrates polymer chains in a matrix.

I am using the same random seed every time I rerun my same input script. However, I am obtaining different values for the velocity and position coordinates in my dump files for every time I repeat the simulation run. Any hints in resolving this issue would be greatly appreciated. Thank you.

​there is not enough information here to give a precise answer. there are *many* ways to produce inconsistent trajectories.
axel.​

 

Sincerely,

Masato Koizumi

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
lammps-users mailing list
lammps-users@...655....net
https://lists.sourceforge.net/lists/listinfo/lammps-users




--
Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.




--
Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.



--
Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.




--
Dr. Axel Kohlmeyer  akohlmey@...92......  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.