[Q-e-developers] [Q-e-commits] r11314 - trunk/espresso/Modules

[Paolo Giannozzi]

On Fri, 2015-01-16 at 10:06 +0100, Filippo Spiga wrote:

> Which MPI flavor? How much old? 

the one that runs of my PC (just updated to debian 7.8, I am
a proud troglodyte :-) ):
$mpirun -help
mpirun (Open MPI) 1.4.5

> IMHO, if anyone has a deployment of a MPI library (...) that it is older 
> than two or three years then it is better to upgrade it (and it is not
> rocket science). MPI_THREAD_FUNNELED is the minimum acceptable level of 
> MPI thread-safety to moving toward a MPI+OpenMP world. We should aim to 
> MPI_THREAD_MULTIPLE in the future (despite not all MPI libraries are already 
> ready for this)

Since I cannot test it, and since nobody in big HPC centers has
complained, I have assumed that there is no hurry to change the way 
MPI+OpenMP is initialized (so I can wait until I upgrade my OS). 
If however such a change is deemed necessary, please go ahead.

Weakly related topics: "configure" still takes into account the case of
Intel 9 (!) fortran (I have Intel 12). There are however an "old" and a
"new" configure right now. Shouldn't we move to the "new"?

> F
> 
> On Jan 13, 2015, at 4:57 PM, giannozz <giannozz at qeforge.qe-forge.org> wrote:
> > Author: giannozz
> > Date: 2015-01-13 16:57:31 +0100 (Tue, 13 Jan 2015)
> > New Revision: 11314
> > 
> > Modified:
> >   trunk/espresso/Modules/mp_world.f90
> > Log:
> > Latest commit unfortunately doesn't seem to work on older MPI versions.
> > reverted to previous version while waiting for better ideas,
> > 
> > 
> > Modified: trunk/espresso/Modules/mp_world.f90
> > ===================================================================
> > --- trunk/espresso/Modules/mp_world.f90	2015-01-13 15:15:38 UTC (rev 11313)
> > +++ trunk/espresso/Modules/mp_world.f90	2015-01-13 15:57:31 UTC (rev 11314)
> > @@ -41,9 +41,6 @@
> >     IMPLICIT NONE
> >     INTEGER, INTENT(IN) :: my_world_comm
> >     INTEGER :: ierr
> > -
> > -    INTEGER :: PROVIDED    
> > -    INTEGER :: MPI_THREAD_FUNNELED
> >     !
> >     world_comm = my_world_comm
> >     !
> > @@ -53,11 +50,7 @@
> >     CALL mpi_initialized ( library_mode, ierr)
> >     IF (ierr/=0) CALL mp_stop( 8000 )
> >     IF (.NOT. library_mode ) THEN
> > -#if defined(__OPENMP)
> > -       CALL mpi_init_thread(MPI_THREAD_FUNNELED,PROVIDED,ierr)
> > -#else
> >        CALL mpi_init(ierr)
> > -#endif
> >        IF (ierr/=0) CALL mp_stop( 8001 )
> >     END IF
> > #endif
> > 
> > _______________________________________________
> > Q-e-commits mailing list
> > Q-e-commits at qe-forge.org
> > http://qe-forge.org/mailman/listinfo/q-e-commits
> 
> --
> Mr. Filippo SPIGA, M.Sc.
> http://filippospiga.info ~ skype: filippo.spiga
> 
> «Nobody will drive us out of Cantor's paradise.» ~ David Hilbert
> 
> *****
> Disclaimer: "Please note this message and any attachments are CONFIDENTIAL and may be privileged or otherwise protected from disclosure. The contents are not to be disclosed to anyone other than the addressee. Unauthorized recipients are requested to preserve this confidentiality and to advise the sender immediately of any error in transmission."
> 
> 

-- 
 Paolo Giannozzi, Dept. Chemistry&Physics&Environment, 
 Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
 Phone +39-0432-558216, fax +39-0432-558222