[Pw_forum] mpi

Axel Kohlmeyer axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Thu Oct 28 17:21:20 CEST 2004


On Thu, 28 Oct 2004, Mahmoud Payami wrote:

MP> Dear Paolo,
MP> 

dear mahmoud,

espresso 2.1 compiles (and runs) nicely with several mpi packages.
for user level, tcp/ip based mpi libraries, most problems (after 
successful compilation) fall into two categories: a) the mpi tools
are not able to log into the compute nodes without a password,
b) you have a faulty network (bad driver or hardware).

i always found MPICH somewhat lacking in helping the user to figure out, 
what has been going wrong. you'll only notice failures of type a) only
when the parallel infractructure is accessed for the first time, which 
may be somewhat after you have started your parallel job, and thus 
i keep recommending LAM-MPI instead. here you can use the lamboot 
command and a hostfile to initialize the parallel environment, and
thus see if you can really log into alle nodes automatically without
a password, then (optionally) test it with the tping command and then 
finally run your parallel job with mpirun. lam-mpi does support
reading input files from stdin, btw.

if you want to save some time compiling the lam-mpi package, you
can downloads suitable for many linux c and fortran compiles
from:
http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/cpmd-linux.html#mpi

to use ifort (8.1) you need to run configure like this:
F90=mpiifort F77=mpiifort CC=mpiicc ./configure

if you want/have to use gcc instead of icc as c-compiler use:
F90=mpiifort F77=mpiifort CC=mpicc ./configure

i'd also recommend to replace the -static in the linker line
of the resulting make.sys file with -static-libcxa.

good luck,
	axel.

MP> I tried to do the compiling job with mpich-2.0.711(which is compiled
MP> by ifc8.1. After installing mpi, I used ./configure for espresso which
MP> could detect the parallel environment and "make all" worked well.
MP> However, trying to run pw.x on single or multi processors, I get
MP> unknown error. Could you please let me know which "mpi" ( from where)
MP> and "ifc" give the desired results.

MP> Thanks in advace.
MP> Best regards,
MP> Mahmoud
MP> 

-- 

=======================================================================
Dr. Axel Kohlmeyer   e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum  http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.




More information about the users mailing list