[Pw_forum] 5.0.1 "cannot allocate spsi" in cegterg

Filippo Spiga spiga.filippo at gmail.com
Sat Jan 19 18:24:56 CET 2013


Dear Anton,

On Jan 19, 2013, at 6:05 PM, Антон Литвиненко <tenebrosus.scriptor at gmail.com> wrote:
> Am 19.01.2013 16:35, schrieb Filippo Spiga:
>> I suggest to compile with "--disable-parallel" and avoid to use mpirun 
>> if you plan to run in serial.
> MPI parallelization on SMP CPU gave inferior performance than OpenMP on 
> my systems one so I'd like to avoid it.

You mean "OpenMP-only parallelization gives low performance than 4 MPI without OpenMP"? This is possible, it depends by the case you are running. But it also depends by the multi-thread libraries you are using. You mentioned ATLAS in a previous email. If you have Intel compiler you also have Intel MKL library... so why do you want to use ATLAS? If you do not compile correctly ATLAS performance will be low...

In case of QE-GPU, having 4 MPI on a multi-core workstation means share the GPU among them and split by a factor of the number of the MPI processes the available RAM on the GPU board (in your case 4). I do not know what kind of GPU do you have but you would like to avoid this scenario. My suggestion is to do not use MPI if you want to exploit the GPU capabilities of QE-GPU on a single workstation.

Cheers,
Filippo

(PS: please subscribe http://qe-forge.org/mailman/listinfo/q-e-gpgpu for all QE-GPU stuff :-)
 
--
Mr. Filippo SPIGA, M.Sc.
http://filippospiga.me ~ skype: filippo.spiga

«Nobody will drive us out of Cantor's paradise.» ~ David Hilbert

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20130119/f310cad5/attachment.html>


More information about the users mailing list