[gmx-users] z always small for domain decomposition grid?

[Szilárd Páll]

More dimensions = more complex communication pattern. See
http://pubs.acs.org/doi/abs/10.1021/ct700301q

Note that it's not *always* x-y decomposition first, mdrun will look
at the dimensions of your system too. One case where manually
switching to 3D decomposition can be beneficial is when there is a
huge imbalance due to bonded interactions in the protein; with a 3D
grid these can be "spread out" among ranks better. You can try other
docomposition grids  using -dd nx ny nz.

Cheers,
--
Szilárd


On Thu, Feb 6, 2014 at 3:14 AM, yunshi11 . <yunshi09 at gmail.com> wrote:
> Hi everyone,
>
> For my MD simulations on different number of CPUs (sometimes with GPUs),
> the domain decomposition grid that I got from automatic (not setting -dd)
> domain decomposition is always like:
> ......
> Domain decomposition grid 8 x 6 x 2, separate PME nodes 48
> PME domain decomposition: 8 x 6 x 1
> ......
>
> or in the case of GPUs:
> ......
> Domain decomposition grid 4 x 3 x 1, separate PME nodes 0
> PME domain decomposition: 4 x 3 x 1
> ......
>
> This happens regardless of the systems (all proteins in explicit solvents)
> I am modeling. So why it is never like 6 x 4 x 4 or 3 x 2 x 2? Isn't it
> better that protein atoms be distributed more equally and into more grid
> boxes?
>
> Thanks,
> Yun
> --
> 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.