Re: Low performance in SMwST simulations

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Fri Aug 26 2022 - 03:29:20 CDT

Next message: zeynab hosseini: "Creating a parameter file for SiC nanotube"
Previous message: Mükerrem Tüfekçioğlu: "Low performance in SMwST simulations"
In reply to: Mükerrem Tüfekçioğlu: "Low performance in SMwST simulations"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

Hello Mukerrem, this is probably due to using functions that aren't
parallelized.
https://urldefense.com/v3/__http://colvars.github.io/colvars-refman-namd/colvars-refman-namd.html*sec:colvar_atom_groups_scaling__;Iw!!DZ3fjg!4Xj4gBX0pnb0b1xW2_Nfth1akc7cTXyPHIBaH3p3AP6YE-pc7LykB4TD_CJpWLdwDyXLvA4kNLNIwi24jZb2K9P5yA$

Can you post a link to the documentation/tutorial/etc that you used as
basis to prepare your SMwST input, to know how best to fill the gaps?

Giacomo

-- 
------------------------------------------------------------------------
This message discusses the usage of Colvars, a software library that is
distributed with NAMD, but is independently developed and maintained.
Please see https://urldefense.com/v3/__https://colvars.github.io__;!!DZ3fjg!4Xj4gBX0pnb0b1xW2_Nfth1akc7cTXyPHIBaH3p3AP6YE-pc7LykB4TD_CJpWLdwDyXLvA4kNLNIwi24jZYrPelc4g$   for further information.
------------------------------------------------------------------------
On Thu, Aug 25, 2022 at 10:35 PM Mükerrem Tüfekçioğlu <
mukerremtufekcioglu_at_gmail.com> wrote:
> Hello everyone,
>
> I have been running simulations using the String Method with Swarms of
> Trajectories, and have been getting unexpected low performance when using
> newer processors.
>
> Using 2000 Intel E5-2680v4 cores, I am getting the same number of
> iterations per hour as when I am using 4000 Intel Xeon 6248R cores. We
> suspect this might be due to different MPI modules available for the two
> different systems, but this idea didn’t lead to solutions. The individual
> simulation log files have the correct number of cores allocated for each of
> them, so it’s not the case that the jobs are using the same number of cores.
>
> I don’t know if this is important, but the MD system in question consists
> of ~180,000 atoms, with 50 images and a total of 1000 replicas.
>
> Any help is appreciated, thank you very much.
>
> Mukerrem
>
>
>

Next message: zeynab hosseini: "Creating a parameter file for SiC nanotube"
Previous message: Mükerrem Tüfekçioğlu: "Low performance in SMwST simulations"
In reply to: Mükerrem Tüfekçioğlu: "Low performance in SMwST simulations"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

This archive was generated by hypermail 2.1.6 : Tue Dec 13 2022 - 14:32:44 CST