[mpich-discuss] MPI process binding on SunOS systems

[Balaji, Pavan]

This is not linux specific.  It's following whatever order the OS gives it.  You might want to look at the -map-by option for better controlling the processes in the way you are looking for.

  -- Pavan

> On Apr 1, 2015, at 11:47 AM, Aurélien Larcher <aurelien.larcher at gmail.com> wrote:
> 
> Hi Rajeev,
> thank you for the pointer, actually this page leads me to another question.
> What is exactly the policy for binding by hwthreads ?
> 
> I can see that:
> 
> larcher at na75:> HYDRA_TOPO_DEBUG=1  mpiexec -n 8 -bind-to hwthread /bin/true | sort
> process 0 binding: 1 0 0 0 0 0 0 0 
> process 1 binding: 0 0 0 0 1 0 0 0 
> process 2 binding: 0 1 0 0 0 0 0 0 
> process 3 binding: 0 0 0 0 0 1 0 0 
> process 4 binding: 0 0 1 0 0 0 0 0 
> process 5 binding: 0 0 0 0 0 0 1 0 
> process 6 binding: 0 0 0 1 0 0 0 0 
> process 7 binding: 0 0 0 0 0 0 0 1 
> 
> Which seems unfortunate because since the processor is a quad core with hyperthreading the same physical core will get the first two threads.
> 
> larcher at na75:> psrinfo -pv
> The physical processor has 4 cores and 8 virtual processors (0-7)
>   The core has 2 virtual processors (0 4)
>   The core has 2 virtual processors (1 5)
>   The core has 2 virtual processors (2 6)
>   The core has 2 virtual processors (3 7)
>     x86 (GenuineIntel 206A7 family 6 model 42 step 7 clock 3392 MHz)
>       Intel(r) Core(tm) i7-2600 CPU @ 3.40GHz
> 
> 
> This is confirmed by:
> 
> HYDRA_TOPO_DEBUG=1  mpiexec -n 4 -bind-to hwthread /bin/true | sort
> process 0 binding: 1 0 0 0 0 0 0 0 
> process 1 binding: 0 0 0 0 1 0 0 0 
> process 2 binding: 0 1 0 0 0 0 0 0 
> process 3 binding: 0 0 0 0 0 1 0 0 
> 
> Is it due to a Linux-specific numbering of logical cores ?
> 
> Thank you,
> 
> Aurelien
> 
> 
> 
> 
> On Wed, Apr 1, 2015 at 6:18 PM, Thakur, Rajeev <thakur at mcs.anl.gov> wrote:
> Yes, see https://wiki.mpich.org/mpich/index.php/Using_the_Hydra_Process_Manager#Process-core_Binding
> 
> Rajeev
> 
> 
> On Apr 1, 2015, at 11:14 AM, Aurélien Larcher <aurelien.larcher at gmail.com>
>  wrote:
> 
> > Hi,
> > running parallel code on my workstation (with OpenIndiana) I discovered that MPI processes were migrated on a regular basis from one core to another.
> >
> > On SunOS systems the command 'pbind' allows binding a process to a specific core.
> > So I am wondering if binding automatically each process to the next available core is supported directly from MPICH ?
> >
> > Best regards
> >
> > Aurelien
> >
> > --
> > -------------------------------------------------------------------------------
> > LARCHER Aurélien          | KTH, School of Computer Science and Communication
> > Work: +46 (0) 8 790 71 42 | Lindstedtsvägen 5, Plan 4, 100 44 Stockholm, SWEDEN
> > -------------------------------------------------------------------------------
> > _______________________________________________
> > discuss mailing list     discuss at mpich.org
> > To manage subscription options or unsubscribe:
> > https://lists.mpich.org/mailman/listinfo/discuss
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
> 
> 
> 
> -- 
> -------------------------------------------------------------------------------
> LARCHER Aurélien          | KTH, School of Computer Science and Communication
> Work: +46 (0) 8 790 71 42 | Lindstedtsvägen 5, Plan 4, 100 44 Stockholm, SWEDEN
> -------------------------------------------------------------------------------
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss

--
Pavan Balaji  ✉️
http://www.mcs.anl.gov/~balaji

_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss