[MPICH] Problem with mpiexec after creating a ring in Windows/Cygwin

Mauro Sgroi maurosgroi at yahoo.it
Thu Jul 5 08:07:30 CDT 2007


Dear all,
I'm new to MPICH2. I compiled the program Abinit using
the MPICH2 library. The compilation of MPICH2 was
fine. I created a ring between two Windows PC (XP and
Win 2000) having the Cygwin tools installed. I not
installed ssh, so I created the ring doing:

mpd --listenport=2500 & on the first PC

and 

mpd --host=C0PC0167 --port=2500 --listenport=2501

Cheching the ring I get:

mpdtrace -l

C0PC0167_2500 (151.91.212.63)
C0PC0165_2501 (151.91.212.129)

Now I'm trying to launch my parallel code on the 2
monoprocessor PC:

mpiexec  -n 2 /usr/local/abinit/5.3/bin/abinip.exe <
tparal_1.files

I get the error:
[unset]: Unable to get AF_INET socket
[unset]: Unable to connect to 151.91.212.129 on 1385
[unset]: aborting job:
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(247): Initialization failed
MPID_Init(71)........: channel initialization failed
MPID_Init(274).......: PMI_Init returned -1

mpiexec is accessing to the second PC on the wrong
port: 1385 instead of 2501.
I tried also to use a machinefile but with no good
results. How can I force mpiexec to look in the good
port?

Thanks a lot and best regards,
Mauro Sgroi.
Italy.




      ___________________________________ 
L'email della prossima generazione? Puoi averla con la nuova Yahoo! Mail: http://it.docs.yahoo.com/nowyoucan.html




More information about the mpich-discuss mailing list