[gmx-users] Slow gmx tools

[Mark Abraham]

Hi,

On Sat, Oct 14, 2017 at 6:25 PM Morten <morten at slyngborg.com> wrote:

>
> Hi, I am running gromacs v. 2016.2. I noticed there are no way of using
> all my cpu power when doing analysis of trajectories. In the older
> version 5.x I could specify that I would do the analysis with only one
> core, which used to be at least 10 times faster.


Very few of the tools have ever had any form of parallelism, and those that
had it haven't had it removed AFAICR. So if you're suggesting there's a
problem somewhere, then a sample command line would be a useful thing to
share. :-)


> This is not possible in
> v 2016.2?


Speaking generally of the tools, it has never been possible.


> It seems very troublesome when ypu have 200 trajectories over
> 1 microsecond each! Any ideas how to increase performance?
>

You can run each trajectory on its own core (probably needing to take care
to use numactl or taskset to keep them apart and on their own core). But
your performance will likely be limited by how effective your hardware is
at reading the data from the disk files. Most analysis tasks are limited by
the rate at which I/O happens, not the computation done with it once it's
read in.

Mark


>
> Best regards
>
> Morten
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>