[CP2K:3802] Re: Request suggestions for cray xe6

carlo antonio pignedoli c.pig... at gmail.com
Thu Apr 26 16:41:23 UTC 2012


Thanks a lot Axel,
I successfully compiled the OpenMP version with gfortran,
I will countercheck with the specialists that I am doing everything right
with the aprun flags

Ciao

Carlo



On Thu, Apr 26, 2012 at 6:36 PM, Axel <akoh... at gmail.com> wrote:
>
>
> On Thursday, April 26, 2012 5:30:23 AM UTC-4, c.pignedoli wrote:
>>
>> Dear all,
>> I have the impression I am missing many recent improvements of the code,
>> I was not able to follow them.
>>
>> I usually run  standard DFT calculations for bulk/surface systems with
>> 700-1500 atoms (10 to 18 electrons each)
>> on a Cray Xe6 (AMD Interlagos 2 x 16-core 64-bit CPUs, 32 GB per compute
>> node)
>>
>> Do you have suggestions for parameters  that would allow for an optimal
>> use of
>> this architecture? (a problem I have is the limited amount of memory
>> per core 1GB)
>
>
> you should compile with OpenMP support (which also means,
> that you have to ditch the crappy default PGI programming environment
> and use gcc/gfortran instead).
>
> for plain quickstep calculations there should be a
> significant performance increase when running with
> 4 MPI tasks per node and 8 threads per MPI task.
>
> due to the massively reduced need for data
> replication, you should not run out of memory.
>
> since interlagos CPUs do some kind of improved
> hyperthreading and two tasks will have to share
> a floating point unit, i would expect that you should
> get a large improvement already by just using
> only half the cores when using plain MPI only.
> make sure that you use the proper "distance"
> flag to aprun, however, since you need to make
> sure you have exactly one MPI task per
> "bulldozer" subunit.
>
> i would talk to the user support specialists
> of the machine, to make sure you get it right.
>
> cheers,
>     axel.
>
>>
>>
>> Thanks in advance for your help
>>
>> Carlo
>
> --
> You received this message because you are subscribed to the Google Groups
> "cp2k" group.
> To view this discussion on the web visit
> https://groups.google.com/d/msg/cp2k/-/wahIvA7eXokJ.
> To post to this group, send email to cp... at googlegroups.com.
> To unsubscribe from this group, send email to
> cp2k+uns... at googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/cp2k?hl=en.



More information about the CP2K-user mailing list