Re: [OMPI users] Compiling C++ program

2006-04-17 Thread Shekhar Tyagi
Hey Brian

I tried your command but i guess its not working, there is a warning and then
nothing much happens, the command window is shown below

 [styagi@borelli Mpi]$ mpiCC -o mathbyvec.out mathbyvec.cc -showme
pgCC-Warning-Unknown switch: -showme

Please help
Shekhar




Quoting Brian Barrett :

> On Apr 16, 2006, at 11:52 AM, Shekhar Tyagi wrote:
>
> > I am new to MPI and prallel programming, recently i made two
> > programs one in C
> > and other in C++. The cluster on which i work is able to compile
> > and execute
> > the C program but its not able to make an executable file for C++
> > program.
> > The command i am using is mpiCC for C++ program but it looks as
> > though, the
> > wrapping compiler sends this to an infinte loop as its not able to
> > form an
> > executable file. My command window looks like this
> >
> > [styagi@borelli Mpi]$ mpiCC -o mathbyvec.out mathbyvec.cc
> >
> > and nothing happens afterwards. Please help me out in this regard.
>
> I'd be surprised if this was the wrapper compiler, but it can be
> pretty easy to choke a C++ compiler and make it take a long time to
> compile.  An easy way to figure out which is the problem is to use
> the command:
>
>mpiCC -o mathbyvec.out mathbyvec.cc -showme
>
> which will cause the wrapper compiler to show what command it is
> going to execute rather than executing it.  You can then take that
> command and run it.  If the underlying compiler takes a long time to
> run, you know where the problem lies.
>
>
> Brian
>
> --
>Brian Barrett
>Open MPI developer
>http://www.open-mpi.org/
>
>
> ___
> users mailing list
> us...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/users
>





Re: [OMPI users] Compiling C++ program

2006-04-17 Thread Brian Barrett

On Apr 17, 2006, at 4:42 AM, Shekhar Tyagi wrote:

I tried your command but i guess its not working, there is a  
warning and then

nothing much happens, the command window is shown below

 [styagi@borelli Mpi]$ mpiCC -o mathbyvec.out mathbyvec.cc -showme
pgCC-Warning-Unknown switch: -showme


Are you sure that the mpiCC you are running is coming from Open MPI?   
The -showme flag should be supported by all wrapper compilers used by  
Open MPI, leading me to think that perhaps you're picking up the  
MPICH wrapper compilers ;).



Brian

--
  Brian Barrett
  Open MPI developer
  http://www.open-mpi.org/




Re: [OMPI users] Error in configuration for crosscompiler PowerPC

2006-04-17 Thread Jorge Parra

Hi,

I downloaded the nightly release version 1.0.3a1r9650 and attempted to 
configure it disabling the fortran bindings for f77 and f90. However 
it gives me a new error when checking for epoll_ctlon broken Linux 
distros:


configure:error: cannot run test program while cross compiling.

Any inputs will be greatly appreciated. I included the outputs of the 
configuration as an attachment.


Thank you,

Jorge



On Fri, 14 Apr 2006, Brian Barrett wrote:


On Apr 12, 2006, at 7:19 PM, Jorge Parra wrote:


I am trying to configure openmpi-1.0.2. However it gives me an
error while running the configuration step. It stops when
trying to determine the "aligment of bool". I attached the outputs
as it is suggested.

The system I am using to build openmpi is an intel P4 running
RedHat Enterprise 3. I am using it to crosscompile for a
system running a power pc 405, and montavista linux 3.1 over it.


It appears that this is an issue with our configure tests with more
recent versions of GCC and cross-compiling (the error will not occur
when not cross compiling).  Fixes have been applied for all the
upcoming releases.  You can grab a nightly release from the 1.0
branch tomorrow and it should work.  The nightly tarballs can be
downloaded from:

  http://www.open-mpi.org/nightly/v1.0/

Releases from the 1.0 series will not properly cross-compile the
Fortran bindings - they must explicitly be disabled with --disable-
mpi-f77 in order for the configure script to complete.  ROMIO on the
1.0 series has also not been tested to make sure it properly cross-
compiles.  This has been fixed for the upcoming 1.1 release, although
you will have to run an external program on the target platform in
order to generate a seed file for some of the data we need in order
to cross-compile our Fortran bindings.  We don't have documentation
on this just yet, but I can provide you with instructions if you need
Fortran bindings.


Hope this helps,

Brian



ompi-output.tar.gz
Description: GNU Zip compressed data