[Pw_forum] Fwd: Mixed MPI-OpenMP

2009-07-24 Thread Paolo Giannozzi

On Jul 24, 2009, at 18:57 , Paolo Giannozzi wrote:

>DFLAGS=...-D__OPENMP _D__FFTW (or -D__ESSL if you have an IBM
>machine with ESSL and AIX, -D__ESSL_LINUX with ESSL and Linux)
>
>compilation flags for openmp (-openmp for ifort)
>
>openMP-aware libraries (essl, mkl, acml I think)
>
>OMP_NUM_THREADS set to the number of desired processes in the
>script that submits the job


correction:
DFLAGS=...-D__OPENMP _D__FFTW + ifort -openmp + mkl
is tested and safe; other combinations may still run into trouble
due to conflicts between "internal" OpenMP and autothreading
libraries.

P.
---
Paolo Giannozzi, Dept of Physics, University of Udine
via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222





[Pw_forum] Fwd: Mixed MPI-OpenMP

2009-07-24 Thread Paolo Giannozzi
Eduardo Ariel Menendez Proupin wrote:

> 1) How should one compile and run? I could not find it on the manual.

because it is not (yet) there: the mixed mpi-openmp parallelization
is still for the brave. If you feel brave:

   DFLAGS=...-D__OPENMP _D__FFTW (or -D__ESSL if you have an IBM
   machine with ESSL and AIX, -D__ESSL_LINUX with ESSL and Linux)

   compilation flags for openmp (-openmp for ifort)

   openMP-aware libraries (essl, mkl, acml I think)

   OMP_NUM_THREADS set to the number of desired processes in the
   script that submits the job

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy


[Pw_forum] Fwd: Mixed MPI-OpenMP

2009-07-24 Thread Eduardo Ariel Menendez Proupin
Hi,
I read the following in Paolo's announcement of version 4.1

 * mixed openMP-MPI parallelization: both CP and PWscf have been
   demonstrated to run on 65536 processors

1) How should one compile and run? I could not find it on the manual.

2)  Could someone, please, suggest a strategy to mix OpenMP and MPI
efficiently? How many OpenMP threads per node, how many MPI processes per
node? Just not to begin experimenting from the ground. For a test case, I
have a cluster infiniband-connected, where each node have two Quadcore
Xeons, i.e., 8 cores per node.

Thanks,





-- 
Eduardo Menendez
Departamento de Fisica
Facultad de Ciencias
Universidad de Chile
Phone: (56)(2)9787439
URL: http://fisica.ciencias.uchile.cl/~emenendez
-- next part --
An HTML attachment was scrubbed...
URL: 
http://www.democritos.it/pipermail/pw_forum/attachments/20090724/863fd07f/attachment.htm