Charm++ parallel programming system

[Ryan Mokos <ryan.mokos AT gmail.com>]

That's ok.  English is not an easy language. :)  Have you tried online translators, such as Google (http://translate.google.com/)?  I know they don't translate perfectly, but they might help.

From your output, I think this is your charm path:

~/Documentos/charm-6.3-mpi-win32/

Which means this should be your charm bin directory:

~/Documentos/charm-6.3-mpi-win32/bin/

This charm bin directory should be a symbolic link to this directory:

~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/bin/

In this directory, you should have several wrappers for AMPI, such as ampicc, ampiCC, etc.  If you don't, let us know.  When executed, these wrappers call the mpi compiler on your system.  Since you want to use mpicc, you should use ampicc.  One way to do this is:

labadm@PSGLABPES10011:~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/tests/ampi/migration$ ~/Documentos/charm-6.3-mpi-win32/bin/ampicc -o migration test.c -tracemode projections

Another way is:

labadm@PSGLABPES10011:~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/tests/ampi/migration$ ../../../../bin/ampicc -o migration test.c -tracemode projections

Ryan

On Thu, Sep 8, 2011 at 9:24 AM, Márcia Silva <marcianasare AT gmail.com> wrote:

So I need to appoint the correct path to mpicc appoint for char/bin/ampicc? Is it? 

Sorry but I can't understand English very well, I am having difficulties to understand that what you want that I do.

Thank you.

On Thu, Sep 8, 2011 at 11:16 AM, Ryan Mokos <ryan.mokos AT gmail.com> wrote:

Hi Márcia,

As Phil said before:

>> You need to pass that option to charmc or the mpicc wrapper that we
>> provide as part of AMPI. That's not something native compilers
>> understand.

The AMPI wrapper for mpicc is charm/bin/ampicc (others include ampiCC, ampicxx, ampif90, etc.).

Ryan

On Thu, Sep 8, 2011 at 7:27 AM, Márcia Silva <marcianasare AT gmail.com> wrote:

First occurred this problem:

labadm@PSGLABPES10011:~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/tests/ampi/migration$ mpicc -o migration test.c -tracemode projections

gcc: projections: Arquivo ou diretório não encontrado

gcc: unrecognized option '-tracemode'

So I copied projections to directory and occurred the problem bellow:

labadm@PSGLABPES10011:~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/tests/ampi/migration$ mpicc -o migration test.c -tracemode projections

gcc: unrecognized option '-tracemode'

projections: file not recognized: File format not recognized

collect2: ld returned 1 exit status

On Sat, Sep 3, 2011 at 2:27 PM, Phil Miller <mille121 AT illinois.edu> wrote:

Could you run 'which mpicc' please?

On Fri, Sep 2, 2011 at 07:50, Márcia Silva <marcianasare AT gmail.com> wrote:
> But I am using mpicc, look that I am doing with code migration:
> labadm@PSGLABPES10011:~/Documentos/charm-6.3-mpi-win32/mpi-linux-bigemulator/tests/ampi/migration$
> mpicc -o migration test.C -tracemode projections
> gcc: projections: Arquivo ou diretório não encontrado
> gcc: unrecognized option '-tracemode'
> test.C:8:37: error: charm.h: Arquivo ou diretório não encontrado
> test.C: In function ‘int main(int, char**)’:
> test.C:18: error: ‘CkPrintf’ was not declared in this scope
> test.C:33: error: ‘CkNumPes’ was not declared in this scope
> test.C:35: error: ‘CkMyPe’ was not declared in this scope
> test.C:39: error: ‘CkAssert’ was not declared in this scope
> test.C:46: error: ‘AMPI_Migrateto’ was not declared in this scope
> What is wrong? With charmc occured the same problem.
> Thankyou
> On Fri, Sep 2, 2011 at 11:39 AM, Phil Miller <mille121 AT illinois.edu> wrote:
>>
>> On Fri, Sep 2, 2011 at 07:35, Márcia Silva <marcianasare AT gmail.com> wrote:
>> > When I use -tracemode projections, occur:
>> > gcc: projections: Arquivo ou diretório não encontrado
>> > gcc: unrecognized option '-tracemode'
>>
>> You need to pass that option to charmc or the mpicc wrapper that we
>> provide as part of AMPI. That's not something native compilers
>> understand.
>>
>> > When I use ./charmrun with +bglog I can generate the file .log, but
>> > Projections need of files sts for show graphics right? So, how I will
>> > use
>> > this files?
>>
>> The +bglog option relates to BigSim emulator traces. Those are a
>> different sort of log file entirely.
>>
>> When you get your program linked with Projections, it will
>> automatically generate Projections log files when you run it.
>>
>>
>> > Thank you.
>> >
>> > On Fri, Sep 2, 2011 at 11:21 AM, Phil Miller <mille121 AT illinois.edu>
>> > wrote:
>> >>
>> >> On Fri, Sep 2, 2011 at 04:33, Márcia Silva <marcianasare AT gmail.com>
>> >> wrote:
>> >> > I am trying use Projections but I can't show the files sts. I do a
>> >> > simulation  with IS of benchmark NAS as follow:
>> >> > mpicc -o is is.c
>> >> > But the files aren't generate.
>> >> > What I am doing wrong?
>> >>
>> >> .sts files are generated by running a charm++ or AMPI program linked
>> >> with the option '-tracemode projections'. I can't even tell if you're
>> >> compiling is.c as plain MPI or AMPI from what you show, and you don't
>> >> mention actually executing the binary at all.
>> >>
>> >> Once you have run a program with Projections tracing enabled, the .sts
>> >> file just contains metadata. The actual traces will be in files named
>> >> is.<processor>.log.
>> >>
>> >> Phil
>> >
>> >
>
>

_______________________________________________
charm mailing list
charm AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/charm

_______________________________________________
ppl mailing list
ppl AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/ppl