[gmx-users] problems with GROMACS 4.6.2

[Szilárd Páll]

Just a few minor details:

- You can set the affinities yourself through the job scheduler which
should give nearly identical results compared to the mdrun internal
affinity if you simply assign cores to mdrun threads in a sequential
order (or with an #physical cores stride if you want to use
HyperThreading);

- At low parallelization and without OpenMP (i.e. -ntomp 1) you may
not notice much regression even though threads won't be pinned.

--
Szilárd


On Tue, Jun 4, 2013 at 7:05 PM, Mark Abraham <mark.j.abraham at gmail.com> wrote:
> Hi GROMACS users,
>
> It's come to our attention that some changes we made in 4.6.2 to the way we
> detect low-level hardware attributes sometimes affects mdrun performance.
> Specifically:
> * using real MPI will likely lead to lower performance than 4.6.1, if
> GROMACS managing the setting of thread affinities would have been a good
> thing
> * on Mac OS, mdrun might hang completely
> If the simulation runs, then it's correct :-)
>
> We've identified fixes and will release 4.6.3 in the next few days. In the
> meantime, we suggest not bothering to install 4.6.2.
>
> Apologies,
>
> 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/Support/Mailing_Lists/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/Support/Mailing_Lists