On Mon, Feb 18, 2013 at 4:31 PM, Josh Milthorpe
<josh.miltho...@anu.edu.au>wrote:
> When -x10rt mpi is specified, X10 uses 'mpicxx' as the post compiler.
> So to run the resulting executable, whatever is in LD_LIBRARY_PATH must be
> consistent with the libraries linked by mpicxx.
>
> You could specify a particular version of MPI to use as the post compiler
> i.e.
>
> $ x10c++ ... -post '/path/to/openmpi/bin/mpicxx # #
> -I/path/to/openmpi/include/openmpi/ompi # -L/path/to/openmpi/lib ...'
>
> (see
> http://x10-lang.org/documentation/practical-x10-programming/native-code-integration.html)
> ... but this is probably overcomplicating it. Just make sure that your
> environment is set up so that 'mpicxx' is the version you want.
>
> I just changed mpicxx and it compiled w/o warnings. On running there were
warnings; for which I used
http://www.rochester.edu/college/gradstudents/jolmsted/blog/2012/09/30/openmpi-warning/
Then I did:
export X10_NPLACES=16
mpirun -np 8 -mca btl ^openib ./montepiobj
(Ignore -mca btl ^openib, which is suppress the warnings)
Output is:
I am running at 8 places
The value of pi is 3.144
So, it has rightly taken 8 places! Thanks a lot.
Can you compile and run a simple MPI program using the same environment
> (including LD_LIBRARY_PATH) as you are using for X10?
>
> Cheers,
>
> Josh
>
>
> On 19/02/13 09:04, Sparsh Mittal wrote:
>
> Thanks a lot. On my cluster, there are two mpi installations. The default
> is old and what I installed is newer.
>
> On compiling I get an error:
>
> x10c++ -x10rt mpi MontePiCluster.x10 -o montepiobj -O
> x10c++: /usr/bin/ld: warning: libmpi_cxx.so.1, needed by
> ~/X10Source/x10.dist/lib/*libx10rt_mpi.so, not found* (try using -rpath
> or -rpath-link)
> /usr/bin/ld: warning: libmpi.so.1, needed by
> ~/X10Source/x10.dist/lib/libx10rt_mpi.so, not found (try using -rpath or
> -rpath-link)
> ~/X10Source/x10.dist/lib/libx10rt_mpi.so: undefined reference to
> `MPI_Op_commutative'
> ~/X10Source/x10.dist/lib/libx10rt_mpi.so: undefined reference to
> `MPI_Reduce_local'
> collect2: ld returned 1 exit status
> x10c++: Non-zero return code: 1
>
> So I provided the path of newer installation using LD_LIBRARY_PATH. It
> compiles but gives warning:
>
> x10c++: /usr/bin/ld: warning: libmpi_cxx.so.1, needed by
> ~/X10Source/x10.dist/lib/libx10rt_mpi.so, may conflict with libmpi_cxx.so.0
> /usr/bin/ld: warning: libmpi.so.1, needed by
> ~/X10Source/x10.dist/lib/libx10rt_mpi.so, may conflict with libmpi.so.0
>
> Moreover, on running it crashes!
>
> My question is: can we specify the location of MPI from which X10 should
> connect, at the time we compile X10? I tried to see but could not
> ascertain. I had compiled X10 as:
>
> cd x10.dist && ant dist -DX10RT_MPI=true
>
> Please let me know. Thanks a lot for your help.
> Thanks and Regards
> Sparsh Mittal
>
>
>
> On Mon, Feb 18, 2013 at 2:27 PM, Josh Milthorpe <josh.miltho...@anu.edu.au
> > wrote:
>
>> Hi Sparsh,
>>
>> to use MPI you must compile using the MPI version of X10RT.
>>
>> $ *x10c++ **-x10rt mpi MontePiCluster.x10 -o montepiobj -O*
>>
>>
>> http://x10-lang.org/documentation/practical-x10-programming/x10rt-implementations.html
>>
>> Cheers,
>>
>> Josh
>>
>>
>> On 19/02/13 04:22, Sparsh Mittal wrote:
>>
>> Hello
>>
>> I am trying to run
>> http://x10-lang.org/documentation/code-examples/small-examples.html?id=103example
>> with MPI, as follows:
>>
>> *export X10_NPLACES=2
>> x10c++ **MontePiCluster.x10 -o montepiobj -O
>> mpirun -np 4 ./montepiobj 1000 2 4
>> *
>> Here ./montepiobj 1000 2 4 tells to run with 1000 guesses, at 2 places
>> with 4 threads.
>>
>> In the MontePiCluster.x10 file, I have just added one statement:
>>
>> *Console.OUT.println(" I am running at "+ places + " places ");*
>>
>> So, the output comes:
>>
>> *I am running at 2 places
>> I am running at 2 places
>> The value of pi is 3.112
>> I am running at 2 places
>> The value of pi is 3.112
>> The value of pi is 3.112
>> I am running at 2 places
>> The value of pi is 3.112*
>>
>>
>> So, my expectation was that with *-np 4,* and* *X10*_*NPLACES=2*, *it
>> will make it 4*2 = 8 places; but what is happening right now is that same
>> m1 is being run at 4 mpi processors. How can we use the mpi processor so
>> that value of pi is calculated at 8 places (only then using MPI is of any
>> use ) and not at 2 places, repeated at 4 processors.
>>
>> I will be thankful for your answers.
>>
>> Thanks and Regards
>> Sparsh Mittal
>>
>>
>>
>>
>> ------------------------------------------------------------------------------
>> The Go Parallel Website, sponsored by Intel - in partnership with Geeknet,
>> is your hub for all things parallel software development, from weekly thought
>> leadership blogs to news, videos, case studies, tutorials, tech docs,
>> whitepapers, evaluation guides, and opinion stories. Check out the most
>> recent posts - join the conversation now. http://goparallel.sourceforge.net/
>>
>>
>>
>> _______________________________________________
>> X10-users mailing
>> listX10-users@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/x10-users
>>
>>
>>
>>
>> ------------------------------------------------------------------------------
>> The Go Parallel Website, sponsored by Intel - in partnership with Geeknet,
>> is your hub for all things parallel software development, from weekly
>> thought
>> leadership blogs to news, videos, case studies, tutorials, tech docs,
>> whitepapers, evaluation guides, and opinion stories. Check out the most
>> recent posts - join the conversation now.
>> http://goparallel.sourceforge.net/
>> _______________________________________________
>> X10-users mailing list
>> X10-users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/x10-users
>>
>>
>
>
> ------------------------------------------------------------------------------
> The Go Parallel Website, sponsored by Intel - in partnership with Geeknet,
> is your hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials, tech docs,
> whitepapers, evaluation guides, and opinion stories. Check out the most
> recent posts - join the conversation now. http://goparallel.sourceforge.net/
>
>
>
> _______________________________________________
> X10-users mailing
> listX10-users@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/x10-users
>
>
>
>
> ------------------------------------------------------------------------------
> The Go Parallel Website, sponsored by Intel - in partnership with Geeknet,
> is your hub for all things parallel software development, from weekly
> thought
> leadership blogs to news, videos, case studies, tutorials, tech docs,
> whitepapers, evaluation guides, and opinion stories. Check out the most
> recent posts - join the conversation now.
> http://goparallel.sourceforge.net/
> _______________________________________________
> X10-users mailing list
> X10-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/x10-users
>
>
------------------------------------------------------------------------------
The Go Parallel Website, sponsored by Intel - in partnership with Geeknet,
is your hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials, tech docs,
whitepapers, evaluation guides, and opinion stories. Check out the most
recent posts - join the conversation now. http://goparallel.sourceforge.net/
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
