[Wien] MPI parallel execution take much time than nonparallel on the same case on the multiprosesor single machine

[hassan jamnejad]

Dear WIEN users,


My system has 4cores  and I use open MPI 1.2.8-17.4. In addition, I try to
use all processors of a single machine. And I think there must be
differences between non-MPI usage of a single processor in a quad system
and MPI usage of all processor in the same machine.

Also is there any way to use all processors of a machine in a non-MPI run?
And is it sufficient to use three quad core to see the difference between
two run (MPI and non-MPI on the same case)?
Thanks.


---------- Forwarded message ----------
From: Laurence Marks <L-marks at northwestern.edu>
To: A Mailing list for WIEN2k users <wien at zeus.theochem.tuwien.ac.at>
Cc:
Date: Thu, 26 Apr 2012 11:22:27 -0500
Subject: Re: [Wien] MPI parallel execution take much time than nonparallel
on the same case on the multiprosesor single machine
How many cores do you have, and what version of mpi are you using?
Running mpi with only 3 processes on one machine is almost certainly
not going to be efficient and for that just stay with non-mpi. With a
dual quadcore (8 cores) or more it can be, provided that the mpi
version you use optimizes well.

N.B., for mpi you want OMP_NUM_THREADS=1, threads and mpi conflict.

On Thu, Apr 26, 2012 at 11:11 AM, jam <hassan.jamnejad at gmail.com> wrote:
> dear all
> recently I switch to MPI parallel execution.I compile the
> scalapack.2.0.1 and GotoBLAS2 and recompiled wien2k sources.bellow you
> will see the compiler option which I set for compiling the wien2k.
>
> Current settings:
>  O   Compiler options:        -ffree-form -O2 -ffree-line-length-none
>  L   Linker Flags:            $(FOPT) -L/home/GotoBLAS2 -lpthread
>  P   Preprocessor flags       '-DParallel'
>  R   R_LIB (LAPACK+BLAS):     -lgoto2 -lpthread
>
>
> Current parallel compilation option settings:
>     RP  RP_LIB(SCALAPACK+PBLAS): -lgoto2 -L/home/scalapack -lscalapack
-lmpi -ldfftw_mpi -ldfftw
>     FP  FPOPT(par.comp.options): -ffree-form -O2
>     MP  MPIRUN commando        : mpirun -np _NP_ -machinefile _HOSTS_
_EXEC_
>
> unfortunately upon execution on the same test case file the cycles
execution fall in long time execution on lapw1para.
> the .machines file is as bellow too:
> 1:localhost:3
> granularity:1
> extrafine:1
>
> also I have set OMP_NUM_THREADS to 4 in .bashrc and the OS istalled on
the machine is openSUSE-11.4.
>
> I would appreciate if anyone can help me on the issue.
> kind regard
> hassan
>
> ______________________________
_________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien



--
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what
nobody else has thought"
Albert Szent-Gyorgi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20120430/6a45584d/attachment.htm>