[mpich-discuss] mpich2-1.3rc2.tar.tar under cygwin+WINDOWS Server 2003 R2 installation problem

[Victor Gillette]

Sorry to bother again, I would like to add this new output to my previous
e-mail; perhaps it is helpful for you to see tis output

vhgillette at TRISO /cygdrive/e/Victor/MCNP/CHECK/VIC/test
$ mpiexec -n 4 -localonly Hi2c.exe
 argx  2280624   argv  ┤
[unset]: unable to decode hostport from 3213
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(385):
MPID_Init(103).......: channel initialization failed
MPID_Init(352).......: PMI_Init returned -1
 argx  2280624   argv  ┤
[unset]: unable to decode hostport from 3206
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(385):
MPID_Init(103).......: channel initialization failed
MPID_Init(352).......: PMI_Init returned -1
 argx  2280624   argv  ┤
[unset]: unable to decode hostport from 3220
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(385):
MPID_Init(103).......: channel initialization failed
MPID_Init(352).......: PMI_Init returned -1
 argx  2280624   argv  ┤
[unset]: unable to decode hostport from 3227
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(385):
MPID_Init(103).......: channel initialization failed
MPID_Init(352).......: PMI_Init returned -1

Thanks

Victor.

On Tue, Oct 12, 2010 at 11:30 AM, Victor Gillette
<victor.gillette at ncsu.edu>wrote:

> Hi. I am interested in running MCNP parallel on a Windows Server 2003 R2 64
> bits System with Intel fortran Compiler and cygwin installed and working.
>
> First I tried installing
>
> mpich2-1.2.1p1-win-x86-64.msi
>
> but mpiexec answered something like smpd was not the right version.
>
> Then I ran smpd -install
>
> and when trying to run mpiexec nothing happened.
>
> I read that hydra would be a more robust manager instead of smpd, so I
> downloaded the source files of
>
> mpich2-1.3rc2.tar.tar
>
> because the pages with binaries were not available and I ran
>
> ./configure
> make
> make install
>
> I found some issues while make, like datatypes for Fortran were not defined
> in mpi.h of mpif.h . What I did to patch these problems is copying the
> values for these definitions from the mpif.h and mpi.h files of the
> previous
> version I had installed. I have no idea if this is completely crazy or a
> reasonable assumption, but make ended OK. There was another issue which was
> that a file setbotf.o was not found in binding/f77 but I found the
> file setbotf.obj , so I
> renamed the file to have .o extension and make and make install ended
> successfully.
>
> I was able to compile Hiworld.c but it doesn't run. I was not able to
> compile Hiworld.f (datatypes ended with ")" when something else was
> expected.
>
> I created all MPICH twice the first time with all defaults, which
> produced HYDRA, and a second time with MPD because I found no
> explanation about how to start HYDRA in the manual.
>
> With MPD I defined the file .mpd.conf in my home folder and the file
> mpd.conf in the /etc folder
>
> I started MPD with
>
> mpd --ncpus=8 &
>
> and if I type
>
> mpdtrace
>
> the code answers my system name.
>
> if I type
>
> mpiexec -n 4 ls
>
> the output of ls is shown 4 times on the screen.
>
> I can create a.exe with
>
> mpicc Hi.c
>
> successfully, however these are the outputs of different commands I tried:
>
> $ mpiexec Hic.exe
> missing num_proc flag: -n, -np, -hosts, or -localonly.
> Unable to parse the mpiexec command arguments.
>
> vhgillette at TRISO /cygdrive/e/Victor/MCNP/CHECK/VIC/test
> $ mpiexec -n 2 Hic.exe
>
> vhgillette at TRISO /cygdrive/e/Victor/MCNP/CHECK/VIC/test
> $ ./Hic.exe
>  argx  2280688   argv  ô
> Hello from processor 0 of 1
>
>
> so, please, could anyone ive me some advise about how to run mpi
> successfully on my system?
>
> Thanks
>
> Victor
>



-- 
==========================
Victor Gillette, Ph.D.
Research Associate
Nuclear Reactor Program
Department of Nuclear Engineering
North Carolina State University
P.O. Box 7909
2500 Stinson Dr.
Raleigh, NC 27695-7909
TE 919-515-3302
==========================
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20101012/9ce7c6f8/attachment-0001.htm>