Re: [Boost-mpi] mixing boost serialization with C MPI functions

Subject: Re: [Boost-mpi] mixing boost serialization with C MPI functions
From: Jiaxin Han (hanjiaxin_at_[hidden])
Date: 2015-11-19 11:25:01

Jiaxin

2015-11-19 15:48 GMT+00:00 Alain Miniussi <alain.miniussi_at_oca.eu>:

>
> Hi,
>
> What makes you think the problem is with boost MPI ? the message only
> mention the C API.
>  Apparently the problem occurs during an intra node communication, and is
> likely to be related with the MPI configuration since I do not think that
> Boost.MPI directly allocate shared memory.
>
>
I have tested using the C MPI funcs for the same task and it succeeded
(commented out in the test program). So I attribute it to an issue with our
local boost installation, or more precisely, we probably haven't correctly
configured boost on the server.

> What version are you using ? does the boost/mpi/config.hpp file defines
> BOOST_MPI_HOMOGENEOUS ? (if not, that could increase the size of the
> messages).
>
> Yes it is defined. I am using 1.59.0, with IBM platform MPI.

> That being said, the archive is supposed to handle its memory.
>
> Cheers
>
> Alain
>
>
> On 19/11/2015 16:07, Jiaxin Han wrote:
>
>> Hi,
>>
>> Could anyone point me to an example of mixing boost serialization with C
>> MPI functions?
>>
>> I guess I need to use boost::mpi::oarchive. But how should I initialize
>> the buffer parameter, and what should I pass to MPI_Send after that? More
>> specifically, I am trying to do the following:
>>
>> typedef vector <int> ParticleList_t;
>> #define MSG_LEN 100000
>>    ParticleList_t sendbuf(MSG_LEN, 1), recvbuf(MSG_LEN);
>>     mpi::packed_oarchive::buffer_type
>> buffer(sizeof(sendbuf[0])*sendbuf.size());
>>     mpi::packed_oarchive oa(world, buffer, boost::archive::no_header);
>>     oa & sendbuf;
>>    if(world.rank()==0)
>>      MPI_Send(oa, 1, MPI_PACKED, 1, 0, MPI_COMM_WORLD);
>>
>> Do I have to make sure buffer is big enough or the oarchive will handle
>> the memory automatically? If the former, what is the correct memory size to
>> hold a vector? I guess it should hold not just vec.data(), but also
>> vec.size().
>>
>> And lastly, oa does not appear to be the correct variable to pass to
>> MPI_Send. Then what should I pass to MPI_Send after creating the archive?
>>
>>
>> I am asking because the boost mpi installation on our server appears to
>> have a limitation on the message size. For example, the attached code
>> reports an error as:
>>
>> terminate called after throwing an instance of
>> 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::mpi::exception>
>> >'
>>    what():  MPI_Alloc_mem: MPI_Alloc_mem: Out of "special" (shared) memory
>> MPI Application rank 0 killed before MPI_Finalize() with signal 6
>>
>> This appears to be a problem only with the boost installation on the
>> server. The code runs correctly on my local machine.
>>
>> Many Thanks!
>>
>> Jiaxin
>>
>>
>>
>> _______________________________________________
>> Boost-mpi mailing list
>> Boost-mpi_at_[hidden]
>> http://lists.boost.org/mailman/listinfo.cgi/boost-mpi
>>
>>
>
> --
> ---
> Alain
> _______________________________________________
> Boost-mpi mailing list
> Boost-mpi_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-mpi
>