[mpich-discuss] Using MPICH2 windows binary with PETSc

Jayesh Krishna jayesh at mcs.anl.gov
Mon Apr 28 09:35:35 CDT 2008


 Hi,

>>> ...if I use winmpiexec...it just hang there...

  Try running the application by asking the launcher to run the application
in a separate window (Check the checkbox which says "run in a separate
window")

>>> ... Is this something like emulating 2 or 4 processors...

  When a user requests the job launcher (mpiexec) to run the MPI job/pgm
with say x processes (mpiexec -n x cpi.exe), the MPI library launches x MPI
processes. The OS scheduler schedules the processes according to its
scheduling policy (which on a multiprocessor machine could be to schedule on
multiple processors -- this is always not the case since the scheduling
policy of the OS has to be smarter than that).

>>> ..I tried to run the modified fpi.exe in a microsoft windows cluster
environment but it failed...
  You will have to re-compile your program using the Microsoft Compute
Cluster Pack SDK
(http://www.microsoft.com/downloads/details.aspx?FamilyID=d8462378-2f68-409d
-9cb3-02312bc23bfd&displaylang=en) to run it in the microsoft windows
cluster.

 Let us know if you have any other issues.

Regards,
Jayesh
-----Original Message-----
From: owner-mpich-discuss at mcs.anl.gov
[mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Ben Tay
Sent: Sunday, April 27, 2008 2:57 AM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Using MPICH2 windows binary with PETSc

Hi Jayesh,

I have tested compiling the example fpi. It ran successfully when I use
mpiexec in command prompt. However, if I use winmpiexec, after inputting a
number, it just hang there. I wonder why. I modify the code to let n=100 and
then it ran w/o any problems.

Also, I found that using processes=1,2,4 gave me slightly different results.
Is this something like emulating 2 or 4 processors? So I can use that to
test whether my mpi code is working properly on 4 or 8 processors?

Lastly, I tried to run the modified fpi.exe in a microsoft windows cluster
environment but it failed. Do you have any experience with it?

The PETSc developers have also emailed me regarding the corrections required
to make MPICH2 work with PETSc. I'll give it a try when I'm free

Thank you very much.

Regards.

Jayesh Krishna wrote:
>  Hi,
>
>   
>>> I am not sure MPICH2 for Windows supports Fortran. Did you check?
>>>       
> 	MPICH2 on windows supports fortran. You must use the right fortran 
> library (See the windows developer's guide for details - available at
> http://www.mcs.anl.gov/research/projects/mpich2/) to link your programs.
>
>   
>>> Even if it does, since you installed binaries, it may not support ifort.
>>>       
> 	MPICH2 supports the Intel fortran compiler. You must link your 
> application (PETSc) with fmpich2.lib (available at MPICH2\lib) for 
> fortran support.
>
>    What is the error that you get when you compile PETSc+MPICH2 ?
>    Are you able to compile the sample application fpi.f available in 
> the MPICH2\examples directory ?
>
> (PS: I will also check with the PETSc developers regarding this 
> issue.)
>
> Regards,
> Jayesh
> -----Original Message-----
> From: owner-mpich-discuss at mcs.anl.gov
> [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Tay Wee Beng
> Sent: Thursday, April 24, 2008 7:25 PM
> To: mpich-discuss at mcs.anl.gov
> Subject: [mpich-discuss] Using MPICH2 windows binary with PETSc
>
> Hi,
>
> I'm trying to compile PETSc with MPICH2 in windows cygwin environment. 
> May I know if it's possible?
>
> I was initially using MPICH1 mpich.nt.1.2.5 and there's no problem. 
> However I can't get it to work with MPICH2. I emailed the developers 
> at PETSc and they asked:
>
> 1) I am not sure MPICH2 for Windows supports Fortran. Did you check?
>
> 2) Even if it does, since you installed binaries, it may not support
ifort.
>
> Can you help to answer these questions?
>
> Hope you can help.
>
> Thank you very much.
>
> Regards.
>
>
>
>
>   





More information about the mpich-discuss mailing list