[gmx-users] Performance difference between MPI ranks and OpenMP

[Carsten Kutzner]

On 08 Jan 2015, at 15:51, Ebert Maximilian <m.ebert at umontreal.ca> wrote:

> The reason I am asking is because I want to use two GPUs and 8 CPUs. So for now I have 2 MPI ranks and 4 OpenMP threads. Is there a way to have 8 MPI ranks but only use 2 GPUs? I also tried 8 MPI ranks with -gpu_id 00001111 but it was about the same as 2 MPI ranks with 4 OpenMP.
You did it correctly and it may very well be that the performance difference
is not that large.

Carsten


> 
> Max
> 
> -----Ursprüngliche Nachricht-----
> Von: gromacs.org_gmx-users-bounces at maillist.sys.kth.se [mailto:gromacs.org_gmx-users-bounces at maillist.sys.kth.se] Im Auftrag von Carsten Kutzner
> Gesendet: Donnerstag, 8. Januar 2015 15:48
> An: gmx-users at gromacs.org
> Betreff: Re: [gmx-users] Performance difference between MPI ranks and OpenMP
> 
> 
> On 08 Jan 2015, at 15:38, Ebert Maximilian <m.ebert at umontreal.ca> wrote:
> 
>> Hi list,
>> 
>> I have another question regarding performance. Is there any performance difference if I start a process on a 8 CPU machine with 8 MPI ranks and 1 OpenMP or 4 MPI ranks an 2 OpenMP? Both should use the 8 CPUs right?
> Right, but there is a performance difference (try it out! :) Normally using just one layer of parallelization is fastest (less overhead), i.e.
> using just OpenMP threads or just MPI ranks. However, more than 8 or so OpenMP threads per rank seldom yield optimal performance, so that a hybrid parallelization approach makes sense in such situations.
> 
> Carsten
> 
> 
>> 
>> Thank you very much,
>> 
>> Max
>> --
>> 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.
> 
> 
> --
> Dr. Carsten Kutzner
> Max Planck Institute for Biophysical Chemistry Theoretical and Computational Biophysics Am Fassberg 11, 37077 Goettingen, Germany Tel. +49-551-2012313, Fax: +49-551-2012302 http://www.mpibpc.mpg.de/grubmueller/kutzner
> http://www.mpibpc.mpg.de/grubmueller/sppexa
> 
> --
> 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.
> -- 
> 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.


--
Dr. Carsten Kutzner
Max Planck Institute for Biophysical Chemistry
Theoretical and Computational Biophysics
Am Fassberg 11, 37077 Goettingen, Germany
Tel. +49-551-2012313, Fax: +49-551-2012302
http://www.mpibpc.mpg.de/grubmueller/kutzner
http://www.mpibpc.mpg.de/grubmueller/sppexa