[Pw_forum] use doual core

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Thu Nov 30 18:14:10 CET 2006


On Thu, 30 Nov 2006, Amin Babazadeh wrote:

amin,

AB> Thanks a lot for your answers.but until yet i couldn't solve my
AB> problem.Iconfused about mpi. there are three kind of mpi in the
AB> Internet

there are _many_ more MPI implementations (some obsolete), not counting 
the commercial ones and platform/network specific implementations.

MPI defines the programmers interface, i.e. what each function should
do and what the arguments should look lile.

(lam/mpi,openmpi,mpich) i don't know which one of them i should install.

just one of them. since you have 'only' a single dual core node, it
does not matter much. communication via shared memory is rather fast.
i would recommend using LAM/MPI for convenience (make sure you defined
-D__LAM in your make.sys file).

AB> about making the environment i should say that i am beginner in 
AB> using Linux so i use the ./configure commend and i didn't change 
AB> any path because i don't know how i should do it.

no need to hack the environment, this is only needed, if you have
multiple MPI libraries installed (or a lazy sysadmin).

AB> i will appreciate if answer my question?
AB> 1-which mpi do i need for my job?

(almost) any would do. see above.

AB> 2-how could i change the path?

you don't need to, just follow the installation
instructions.

AB> 3-is ifort compiler enough for installing mpi package?

why not? you just have to make sure that you compile
your mpi package, so that it uses ifort and not another
fortran compiler. see the installation instructions
of your MPI package of choice on how to specify your
compiler preferences. 

please note, that all of this would be more appropriate
to ask at an mpi support mailing list, as these are problems
that are completely independent of quantum espresso.

also, after installation of your mpi package of choice, 
please test it first with some _simple_ mpi programs.
most packages supply a few testcases and examples. pay
special attention to the fortran examples if you care 
about using mpi for quantum espresso. once you have
verified that your mpi installation is working correctly
(and only after that), you should move to trying to
compiler quantum espresso in parallel...

even if off-topic, i hope that there are some people 
around here, that will benefit from those clarifications.

cheers,
   axel.

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.




More information about the users mailing list