[library Boost.MPI
    [authors [Gregor, Douglas], [Troyer, Matthias] ]
    [copyright 2005 2006 2007 Douglas Gregor, Matthias Troyer, Trustees of Indiana University]
    [purpose
        An generic, user-friendly interface to MPI, the Message
        Passing Interface.
    ]
    [id mpi]
    [dirname mpi]
    [license
        Distributed under the Boost Software License, Version 1.0.
        (See accompanying file LICENSE_1_0.txt or copy at
        <ulink url="http://www.boost.org/LICENSE_1_0.txt">
            http://www.boost.org/LICENSE_1_0.txt
        </ulink>)
    ]
]

[/ Links ]
[def _MPI_         [@http://www-unix.mcs.anl.gov/mpi/ MPI]]
[def _MPI_implementations_ 
   [@http://www-unix.mcs.anl.gov/mpi/implementations.html
    MPI implementations]]
[def _Serialization_ [@http://www.boost.org/libs/serialization/doc
                      Boost.Serialization]]
[def _BoostPython_ [@http://www.boost.org/libs/python/doc
                      Boost.Python]]
[def _Python_      [@http://www.python.org Python]]
[def _LAM_          [@http://www.lam-mpi.org/ LAM/MPI]]
[def _MPICH_        [@http://www-unix.mcs.anl.gov/mpi/mpich/ MPICH]]
[def _OpenMPI_      [@http://www.open-mpi.org OpenMPI]]
[def _accumulate_   [@http://www.sgi.com/tech/stl/accumulate.html
                     `accumulate`]]

[/ QuickBook Document version 1.0 ]

[section:intro Introduction]

Boost.MPI is a library for message passing in high-performance
parallel applications. A Boost.MPI program is one or more processes
that can communicate either via sending and receiving individual
messages (point-to-point communication) or by coordinating as a group
(collective communication). Unlike communication in threaded
environments or using a shared-memory library, Boost.MPI processes can
be spread across many different machines, possibly with different
operating systems and underlying architectures. 

Boost.MPI is not a completely new parallel programming
library. Rather, it is a C++-friendly interface to the standard
Message Passing Interface (_MPI_), the most popular library interface
for high-performance, distributed computing. MPI defines
a library interface, available from C, Fortran, and C++, for which
there are many _MPI_implementations_. Although there exist C++
bindings for MPI, they offer little functionality over the C
bindings. The Boost.MPI library provides an alternative C++ interface
to MPI that better supports modern C++ development styles, including
complete support for user-defined data types and C++ Standard Library
types, arbitrary function objects for collective algorithms, and the
use of modern C++ library techniques to maintain maximal
efficiency.

At present, Boost.MPI supports the majority of functionality in MPI
1.1. The thin abstractions in Boost.MPI allow one to easily combine it
with calls to the underlying C MPI library. Boost.MPI currently
supports:

* Communicators: Boost.MPI supports the creation,
  destruction, cloning, and splitting of MPI communicators, along with
  manipulation of process groups. 
* Point-to-point communication: Boost.MPI supports
  point-to-point communication of primitive and user-defined data
  types with send and receive operations, with blocking and
  non-blocking interfaces.
* Collective communication: Boost.MPI supports collective
  operations such as [funcref boost::mpi::reduce `reduce`]
  and [funcref boost::mpi::gather `gather`] with both
  built-in and user-defined data types and function objects.
* MPI Datatypes: Boost.MPI can build MPI data types for
  user-defined types using the _Serialization_ library.
* Separating structure from content: Boost.MPI can transfer the shape
  (or "skeleton") of complexc data structures (lists, maps,
  etc.) and then separately transfer their content. This facility
  optimizes for cases where the data within a large, static data
  structure needs to be transmitted many times.

Boost.MPI can be accessed either through its native C++ bindings, or
through its alternative, [link mpi.python Python interface].

[endsect]

[section:getting_started Getting started]

Getting started with Boost.MPI requires a working MPI implementation,
a recent version of Boost, and some configuration information.

[section:mpi_impl MPI Implementation]
To get started with Boost.MPI, you will first need a working
MPI implementation. There are many conforming _MPI_implementations_
available. Boost.MPI should work with any of the
implementations, although it has only been tested extensively with:

* [@http://www.open-mpi.org Open MPI 1.0.x]
* [@http://www.lam-mpi.org LAM/MPI 7.x]
* [@http://www-unix.mcs.anl.gov/mpi/mpich/ MPICH 1.2.x]

You can test your implementation using the following simple program,
which passes a message from one processor to another. Each processor
prints a message to standard output. 

  #include <mpi.h>
  #include <iostream>

  int main(int argc, char* argv[])
  {
    MPI_Init(&argc, &argv);

    int rank;
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    if (rank == 0) {
      int value = 17;
      int result = MPI_Send(&value, 1, MPI_INT, 1, 0, MPI_COMM_WORLD);
      if (result == MPI_SUCCESS)
        std::cout << "Rank 0 OK!" << std::endl;
    } else if (rank == 1) {
      int value;
      int result = MPI_Recv(&value, 1, MPI_INT, 0, 0, MPI_COMM_WORLD,
  			  MPI_STATUS_IGNORE);
      if (result == MPI_SUCCESS && value == 17)
        std::cout << "Rank 1 OK!" << std::endl;
    }
    MPI_Finalize();
    return 0;
  } 

You should compile and run this program on two processors. To do this,
consult the documentation for your MPI implementation. With _LAM_, for
instance, you compile with the `mpiCC` or `mpic++` compiler, boot the
LAM/MPI daemon, and run your program via `mpirun`. For instance, if
your program is called `mpi-test.cpp`, use the following commands:

[pre
mpiCC -o mpi-test mpi-test.cpp
lamboot
mpirun -np 2 ./mpi-test
lamhalt
]

When you run this program, you will see both `Rank 0 OK!` and `Rank 1
OK!` printed to the screen. However, they may be printed in any order
and may even overlap each other. The following output is perfectly
legitimate for this MPI program:

[pre
Rank Rank 1 OK!
0 OK!
]

If your output looks something like the above, your MPI implementation
appears to be working with a C++ compiler and we're ready to move on.
[endsect]

[section:config Configure and Build]

Boost.MPI uses version 2 of the
[@http://www.boost.org/doc/html/bbv2.html Boost.Build] system for
configuring and building the library binary. You will need a very new
version of [@http://www.boost.org/tools/build/jam_src/index.html
Boost.Jam] (3.1.12 or later). If you already have Boost.Jam, run `bjam
-v` to determine what version you are using.

Information about building Boost.Jam is
[@http://www.boost.org/tools/build/jam_src/index.html#building_bjam
available here]. However, most users need only run `build.sh` in the
`tools/build/jam_src` subdirectory of Boost. Then,
copy the resulting `bjam` executable some place convenient.

For many users using _LAM_, _MPICH_, or _OpenMPI_, configuration is
almost automatic. If you don't already have a file `user-config.jam`
in your home directory, copy `tools/build/v2/user-config.jam`
there. For many users, MPI support can be enabled simply by adding the
following line to your user-config.jam file, which is used to configure
Boost.Build version 2. 

  using mpi ;

This should auto-detect MPI settings based on the MPI wrapper compiler in 
your path, e.g., `mpic++`. If the wrapper compiler is not in your
path, see below.

To actually build the MPI library, go into the top-level Boost
directory and execute the command:

[pre
bjam --with-mpi
]

If your MPI wrapper compiler has a different name from the default,
you can pass the name of the wrapper compiler as the first argument to
the mpi module:

  using mpi : /opt/mpich2-1.0.4/bin/mpiCC ;

If your MPI implementation does not have a wrapper compiler, or the MPI 
auto-detection code does not work with your MPI's wrapper compiler,
you can pass MPI-related options explicitly via the second parameter to the 
`mpi` module:

   using mpi : : <find-shared-library>lammpio <find-shared-library>lammpi++
                 <find-shared-library>mpi <find-shared-library>lam 
                 <find-shared-library>dl ;

To see the results of MPI auto-detection, pass `--debug-configuration` on
the bjam command line.

The (optional) fourth argument configures Boost.MPI for running
regression tests. These parameters specify the executable used to
launch jobs (default: "mpirun") followed by any necessary arguments
to this to run tests and tell the program to expect the number of
processors to follow (default: "-np").  With the default parameters,
for instance, the test harness will execute, e.g.,

[pre  
mpirun -np 4 all_gather_test
]

[endsect]

[section:installation Installing and Using Boost.MPI]

Installation of Boost.MPI can be performed in the build step by
specifying `install` on the command line and (optionally) providing an
installation location, e.g.,

[pre
bjam --with-mpi install
]

This command will install libraries into a default system location. To
change the path where libraries will be installed, add the option
`--prefix=PATH`.

To build applications based on Boost.MPI, compile and link them as you
normally would for MPI programs, but remember to link against the
`boost_mpi` and `boost_serialization` libraries, e.g.,

[pre
mpic++ -I/path/to/boost/mpi my_application.cpp -Llibdir \
  -lboost_mpi-gcc-mt-1_35 -lboost_serialization-gcc-d-1_35.a
]
[endsect]

If you plan to use the [link mpi.python Python bindings] for
Boost.MPI in conjunction with the C++ Boost.MPI, you will also need to
link against the boost_mpi_python library, e.g., by adding
`-lboost_mpi_python-gcc-mt-1_35` to your link command. This step will
only be necessary if you intend to [link mpi.python_user_data
register C++ types] or use the [link
mpi.python_skeleton_content skeleton/content mechanism] from
within Python.

[section:testing Testing Boost.MPI] 

If you would like to verify that Boost.MPI is working properly with
your compiler, platform, and MPI implementation, a self-contained test
suite is available. To use this test suite, you will need to first
configure Boost.Build for your MPI environment and then run `bjam` in
`libs/mpi/test` (possibly with some extra options). For 
_LAM_, you will need to run `lamboot` before running `bjam`. For
_MPICH_, you may need to create a machine file and pass
`-sMPIRUN_FLAGS="-machinefile <filename>"` to Boost.Jam; see the
section on [link mpi.config configuration] for more
information. If testing succeeds, `bjam` will exit without errors.

[endsect]

[endsect]

[section:tutorial Tutorial]

A Boost.MPI program consists of many cooperating processes (possibly
running on different computers) that communicate among themselves by
passing messages. Boost.MPI is a library (as is the lower-level MPI),
not a language, so the first step in a Boost.MPI is to create an
[classref boost::mpi::environment mpi::environment] object
that initializes the MPI environment and enables communication among
the processes. The [classref boost::mpi::environment
mpi::environment] object is initialized with the program arguments
(which it may modify) in your main program. The creation of this
object initializes MPI, and its destruction will finalize MPI. In the
vast majority of Boost.MPI programs, an instance of [classref
boost::mpi::environment mpi::environment] will be declared
in `main` at the very beginning of the program.

Communication with MPI always occurs over a *communicator*,
which can be created be simply default-constructing an object of type
[classref boost::mpi::communicator mpi::communicator]. This
communicator can then be queried to determine how many processes are
running (the "size" of the communicator) and to give a unique number
to each process, from zero to the size of the communicator (i.e., the
"rank" of the process):

  #include <boost/mpi/environment.hpp>
  #include <boost/mpi/communicator.hpp>
  #include <iostream>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[]) 
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;
    std::cout << "I am process " << world.rank() << " of " << world.size()
              << "." << std::endl;
    return 0;
  }

If you run this program with 7 processes, for instance, you will
receive output such as:

[pre
I am process 5 of 7.
I am process 0 of 7.
I am process 1 of 7.
I am process 6 of 7.
I am process 2 of 7.
I am process 4 of 7.
I am process 3 of 7.
]

Of course, the processes can execute in a different order each time,
so the ranks might not be strictly increasing. More interestingly, the
text could come out completely garbled, because one process can start
writing "I am a process" before another process has finished writing
"of 7.".

[section:point_to_point Point-to-Point communication]

As a message passing library, MPI's primary purpose is to routine
messages from one process to another, i.e., point-to-point. MPI
contains routines that can send messages, receive messages, and query
whether messages are available. Each message has a source process, a
target process, a tag, and a payload containing arbitrary data. The
source and target processes are the ranks of the sender and receiver
of the message, respectively. Tags are integers that allow the
receiver to distinguish between different messages coming from the
same sender. 

The following program uses two MPI processes to write "Hello, world!"
to the screen (`hello_world.cpp`):

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <string>
  #include <boost/serialization/string.hpp>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[]) 
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    if (world.rank() == 0) {
      world.send(1, 0, std::string("Hello"));
      std::string msg;
      world.recv(1, 1, msg);
      std::cout << msg << "!" << std::endl;
    } else {
      std::string msg;
      world.recv(0, 0, msg);
      std::cout << msg << ", ";
      std::cout.flush();
      world.send(0, 1, std::string("world"));
    }

    return 0;
  }

The first processor (rank 0) passes the message "Hello" to the second
processor (rank 1) using tag 0. The second processor prints the string
it receives, along with a comma, then passes the message "world" back
to processor 0 with a different tag. The first processor then writes
this message with the "!" and exits. All sends are accomplished with
the [memberref boost::mpi::communicator::send
communicator::send] method and all receives use a corresponding
[memberref boost::mpi::communicator::recv
communicator::recv] call.

[section:nonblocking Non-blocking communication]

The default MPI communication operations--`send` and `recv`--may have
to wait until the entire transmission is completed before they can
return. Sometimes this *blocking* behavior has a negative impact on
performance, because the sender could be performing useful computation
while it is waiting for the transmission to occur. More important,
however, are the cases where several communication operations must
occur simultaneously, e.g., a process will both send and receive at
the same time.

Let's revisit our "Hello, world!" program from the previous
section. The core of this program transmits two messages:

    if (world.rank() == 0) {
      world.send(1, 0, std::string("Hello"));
      std::string msg;
      world.recv(1, 1, msg);
      std::cout << msg << "!" << std::endl;
    } else {
      std::string msg;
      world.recv(0, 0, msg);
      std::cout << msg << ", ";
      std::cout.flush();
      world.send(0, 1, std::string("world"));
    }

The first process passes a message to the second process, then
prepares to receive a message. The second process does the send and
receive in the opposite order. However, this sequence of events is
just that--a *sequence*--meaning that there is essentially no
parallelism. We can use non-blocking communication to ensure that the
two messages are transmitted simultaneously
(`hello_world_nonblocking.cpp`): 

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <string>
  #include <boost/serialization/string.hpp>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[]) 
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    if (world.rank() == 0) {
      mpi::request reqs[2];
      std::string msg, out_msg = "Hello";
      reqs[0] = world.isend(1, 0, out_msg);
      reqs[1] = world.irecv(1, 1, msg);
      mpi::wait_all(reqs, reqs + 2);
      std::cout << msg << "!" << std::endl;
    } else {
      mpi::request reqs[2];
      std::string msg, out_msg = "world";
      reqs[0] = world.isend(0, 1, out_msg);
      reqs[1] = world.irecv(0, 0, msg);
      mpi::wait_all(reqs, reqs + 2);
      std::cout << msg << ", ";
    }

    return 0;
  }

We have replaced calls to the [memberref
boost::mpi::communicator::send communicator::send] and
[memberref boost::mpi::communicator::recv
communicator::recv] members with similar calls to their non-blocking
counterparts, [memberref boost::mpi::communicator::isend
communicator::isend] and [memberref
boost::mpi::communicator::irecv communicator::irecv]. The
prefix *i* indicates that the operations return immediately with a
[classref boost::mpi::request mpi::request] object, which
allows one to query the status of a communication request (see the
[memberref boost::mpi::request::test test] method) or wait
until it has completed (see the [memberref
boost::mpi::request::wait wait] method). Multiple requests
can be completed at the same time with the [funcref
boost::mpi::wait_all wait_all] operation. 

If you run this program multiple times, you may see some strange
results: namely, some runs will produce:

  Hello, world!

while others will produce:

  world!
  Hello,

or even some garbled version of the letters in "Hello" and
"world". This indicates that there is some parallelism in the program,
because after both messages are (simultaneously) transmitted, both
processes will concurrent execute their print statements. For both
performance and correctness, non-blocking communication operations are
critical to many parallel applications using MPI.

[endsect]

[section:user_data_types User-defined data types]

The inclusion of `boost/serialization/string.hpp` in the previous
examples is very important: it makes values of type `std::string`
serializable, so that they can be be transmitted using Boost.MPI. In
general, built-in C++ types (`int`s, `float`s, characters, etc.) can
be transmitted over MPI directly, while user-defined and
library-defined types will need to first be serialized (packed) into a
format that is amenable to transmission. Boost.MPI relies on the
_Serialization_ library to serialize and deserialize data types. 

For types defined by the standard library (such as `std::string` or
`std::vector`) and some types in Boost (such as `boost::variant`), the
_Serialization_ library already contains all of the required
serialization code. In these cases, you need only include the
appropriate header from the `boost/serialization` directory. 

For types that do not already have a serialization header, you will
first need to implement serialization code before the types can be
transmitted using Boost.MPI. Consider a simple class `gps_position`
that contains members `degrees`, `minutes`, and `seconds`. This class
is made serializable by making it a friend of
`boost::serialization::access` and introducing the templated
`serialize()` function, as follows:

  class gps_position
  {
  private:
      friend class boost::serialization::access;

      template<class Archive>
      void serialize(Archive & ar, const unsigned int version)
      {
          ar & degrees;
          ar & minutes;
          ar & seconds;
      }

      int degrees;
      int minutes;
      float seconds;
  public:
      gps_position(){};
      gps_position(int d, int m, float s) :
          degrees(d), minutes(m), seconds(s)
      {}
  };

Complete information about making types serializable is beyond the
scope of this tutorial. For more information, please see the
_Serialization_ library tutorial from which the above example was
extracted. One important side benefit of making types serializable for
Boost.MPI is that they become serializable for any other usage, such
as storing the objects to disk to manipulated them in XML.

Some serializable types, like `gps_position` above, have a fixed
amount of data stored at fixed field positions. When this is the case,
Boost.MPI can optimize their serialization and transmission to avoid
extraneous copy operations. To enable this optimization, users should
specialize the type trait [classref
boost::mpi::is_mpi_datatype `is_mpi_datatype`], e.g.:

  namespace boost { namespace mpi {
    template <>
    struct is_mpi_datatype<gps_position> : mpl::true_ { };
  } }

For non-template types we have defined a macro to simplify declaring a type 
as an MPI datatype

  BOOST_IS_MPI_DATATYPE(gps_position)

For composite traits, the specialization of [classref
boost::mpi::is_mpi_datatype `is_mpi_datatype`] may depend on
`is_mpi_datatype` itself. For instance, a `boost::array` object is
fixed only when the type of the parameter it stores is fixed:

  namespace boost { namespace mpi {
    template <typename T, std::size_t N>
    struct is_mpi_datatype<array<T, N> > 
      : public is_mpi_datatype<T> { };
  } }
  
The redundant copy elimination optimization can only be applied when
the shape of the data type is completely fixed. Variable-length types
(e.g., strings, linked lists) and types that store pointers cannot use
the optimiation, but Boost.MPI will be unable to detect this error at
compile time. Attempting to perform this optimization when it is not
correct will likely result in segmentation faults and other strange
program behavior.

Boost.MPI can transmit any user-defined data type from one process to
another. Built-in types can be transmitted without any extra effort;
library-defined types require the inclusion of a serialization header;
and user-defined types will require the addition of serialization
code. Fixed data types can be optimized for transmission using the
[classref boost::mpi::is_mpi_datatype `is_mpi_datatype`]
type trait.

[endsect]
[endsect]

[section:collectives Collective operations]

[link mpi.point_to_point Point-to-point operations] are the
core message passing primitives in Boost.MPI. However, many
message-passing applications also require higher-level communication
algorithms that combine or summarize the data stored on many different
processes. These algorithms support many common tasks such as
"broadcast this value to all processes", "compute the sum of the
values on all processors" or "find the global minimum." 

[section:broadcast Broadcast]
The [funcref boost::mpi::broadcast `broadcast`] algorithm is
by far the simplest collective operation. It broadcasts a value from a
single process to all other processes within a [classref
boost::mpi::communicator communicator]. For instance, the
following program broadcasts "Hello, World!" from process 0 to every
other process. (`hello_world_broadcast.cpp`)

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <string>
  #include <boost/serialization/string.hpp>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[])
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    std::string value;
    if (world.rank() == 0) {
      value = "Hello, World!";
    }

    broadcast(world, value, 0);

    std::cout << "Process #" << world.rank() << " says " << value 
              << std::endl;
    return 0;
  } 

Running this program with seven processes will produce a result such
as:

[pre
Process #0 says Hello, World!
Process #2 says Hello, World!
Process #1 says Hello, World!
Process #4 says Hello, World!
Process #3 says Hello, World!
Process #5 says Hello, World!
Process #6 says Hello, World!
]
[endsect]

[section:gather Gather]
The [funcref boost::mpi::gather `gather`] collective gathers
the values produced by every process in a communicator into a vector
of values on the "root" process (specified by an argument to
`gather`). The /i/th element in the vector will correspond to the
value gathered fro mthe /i/th process. For instance, in the following
program each process computes its own random number. All of these
random numbers are gathered at process 0 (the "root" in this case),
which prints out the values that correspond to each processor. 
(`random_gather.cpp`)

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <vector>
  #include <cstdlib>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[])
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    std::srand(time(0) + world.rank());
    int my_number = std::rand();
    if (world.rank() == 0) {
      std::vector<int> all_numbers;
      gather(world, my_number, all_numbers, 0);
      for (int proc = 0; proc < world.size(); ++proc)
        std::cout << "Process #" << proc << " thought of " 
                  << all_numbers[proc] << std::endl;
    } else {
      gather(world, my_number, 0);
    }

    return 0;
  } 

Executing this program with seven processes will result in output such
as the following. Although the random values will change from one run
to the next, the order of the processes in the output will remain the
same because only process 0 writes to `std::cout`.

[pre
Process #0 thought of 332199874
Process #1 thought of 20145617
Process #2 thought of 1862420122
Process #3 thought of 480422940
Process #4 thought of 1253380219
Process #5 thought of 949458815
Process #6 thought of 650073868
]

The `gather` operation collects values from every process into a
vector at one process. If instead the values from every process need
to be collected into identical vectors on every process, use the
[funcref boost::mpi::all_gather `all_gather`] algorithm,
which is semantically equivalent to calling `gather` followed by a
`broadcast` of the resulting vector.

[endsect]

[section:reduce Reduce] 

The [funcref boost::mpi::reduce `reduce`] collective
summarizes the values from each process into a single value at the
user-specified "root" process. The Boost.MPI `reduce` operation is
similar in spirit to the STL _accumulate_ operation, because it takes
a sequence of values (one per process) and combines them via a
function object. For instance, we can randomly generate values in each
process and the compute the minimum value over all processes via a
call to [funcref boost::mpi::reduce `reduce`]
(`random_min.cpp`)::

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <cstdlib>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[])
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    std::srand(time(0) + world.rank());
    int my_number = std::rand();

    if (world.rank() == 0) {
      int minimum;
      reduce(world, my_number, minimum, mpi::minimum<int>(), 0);
      std::cout << "The minimum value is " << minimum << std::endl;
    } else {
      reduce(world, my_number, mpi::minimum<int>(), 0);
    }

    return 0;
  }

The use of `mpi::minimum<int>` indicates that the minimum value
should be computed. `mpi::minimum<int>` is a binary function object
that compares its two parameters via `<` and returns the smaller
value. Any associative binary function or function object will
work. For instance, to concatenate strings with `reduce` one could use
the function object `std::plus<std::string>` (`string_cat.cpp`):

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <string>
  #include <functional>
  #include <boost/serialization/string.hpp>
  namespace mpi = boost::mpi;

  int main(int argc, char* argv[])
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    std::string names[10] = { "zero ", "one ", "two ", "three ", 
                              "four ", "five ", "six ", "seven ", 
                              "eight ", "nine " };

    std::string result;
    reduce(world, 
           world.rank() < 10? names[world.rank()] 
                            : std::string("many "),
           result, std::plus<std::string>(), 0);

    if (world.rank() == 0)
      std::cout << "The result is " << result << std::endl;

    return 0;
  } 

In this example, we compute a string for each process and then perform
a reduction that concatenates all of the strings together into one,
long string. Executing this program with seven processors yields the
following output:

[pre
The result is zero one two three four five six
]

Any kind of binary function objects can be used with `reduce`. For
instance, and there are many such function objects in the C++ standard
`<functional>` header and the Boost.MPI header
`<boost/mpi/operations.hpp>`. Or, you can create your own
function object. Function objects used with `reduce` must be
associative, i.e. `f(x, f(y, z))` must be equivalent to `f(f(x, y),
z)`. If they are also commutative (i..e, `f(x, y) == f(y, x)`),
Boost.MPI can use a more efficient implementation of `reduce`. To
state that a function object is commutative, you will need to
specialize the class [classref boost::mpi::is_commutative
`is_commutative`]. For instance, we could modify the previous example
by telling Boost.MPI that string concatenation is commutative:

  namespace boost { namespace mpi {

    template<>
    struct is_commutative<std::plus<std::string>, std::string> 
      : mpl::true_ { };

  } } // end namespace boost::mpi

By adding this code prior to `main()`, Boost.MPI will assume that
string concatenation is commutative and employ a different parallel
algorithm for the `reduce` operation. Using this algorithm, the
program outputs the following when run with seven processes:

[pre
The result is zero one four five six two three
]

Note how the numbers in the resulting string are in a different order:
this is a direct result of Boost.MPI reordering operations. The result
in this case differed from the non-commutative result because string
concatenation is not commutative: `f("x", "y")` is not the same as
`f("y", "x")`, because argument order matters. For truly commutative
operations (e.g., integer addition), the more efficient commutative
algorithm will produce the same result as the non-commutative
algorithm. Boost.MPI also performs direct mappings from function
objects in `<functional>` to `MPI_Op` values predefined by MPI (e.g.,
`MPI_SUM`, `MPI_MAX`); if you have your own function objects that can
take advantage of this mapping, see the class template [classref
boost::mpi::is_mpi_op `is_mpi_op`].

Like [link mpi.gather `gather`], `reduce` has an "all"
variant called [funcref boost::mpi::all_reduce `all_reduce`]
that performs the reduction operation and broadcasts the result to all
processes. This variant is useful, for instance, in establishing
global minimum or maximum values.

[endsect]

[endsect]

[section:communicators Managing communicators]

Communication with Boost.MPI always occurs over a communicator. A
communicator contains a set of processes that can send messages among
themselves and perform collective operations. There can be many
communicators within a single program, each of which contains its own
isolated communication space that acts independently of the other
communicators. 

When the MPI environment is initialized, only the "world" communicator
(called `MPI_COMM_WORLD` in the MPI C and Fortran bindings) is
available. The "world" communicator, accessed by default-constructing
a [classref boost::mpi::communicator mpi::communicator]
object, contains all of the MPI processes present when the program
begins execution. Other communicators can then be constructed by
duplicating or building subsets of the "world" communicator. For
instance, in the following program we split the processes into two
groups: one for processes generating data and the other for processes
that will collect the data. (`generate_collect.cpp`)

  #include <boost/mpi.hpp>
  #include <iostream>
  #include <cstdlib>
  #include <boost/serialization/vector.hpp>
  namespace mpi = boost::mpi;

  enum message_tags {msg_data_packet, msg_broadcast_data, msg_finished};

  void generate_data(mpi::communicator local, mpi::communicator world);
  void collect_data(mpi::communicator local, mpi::communicator world);

  int main(int argc, char* argv[])
  {
    mpi::environment env(argc, argv);
    mpi::communicator world;

    bool is_generator = world.rank() < 2 * world.size() / 3;
    mpi::communicator local = world.split(is_generator? 0 : 1);
    if (is_generator) generate_data(local, world);
    else collect_data(local, world);

    return 0;
  }

When communicators are split in this way, their processes retain
membership in both the original communicator (which is not altered by
the split) and the new communicator. However, the ranks of the
processes may be different from one communicator to the next, because
the rank values within a communicator are always contiguous values
starting at zero. In the example above, the first two thirds of the
processes become "generators" and the remaining processes become
"collectors". The ranks of the "collectors" in the `world`
communicator will be 2/3 `world.size()` and greater, whereas the ranks
of the same collector processes in the `local` communicator will start
at zero. The following excerpt from `collect_data()` (in
`generate_collect.cpp`) illustrates how to manage multiple
communicators:

  mpi::status msg = world.probe();
  if (msg.tag() == msg_data_packet) {
    // Receive the packet of data
    std::vector<int> data;
    world.recv(msg.source(), msg.tag(), data);

    // Tell each of the collectors that we'll be broadcasting some data
    for (int dest = 1; dest < local.size(); ++dest)
      local.send(dest, msg_broadcast_data, msg.source());

    // Broadcast the actual data.
    broadcast(local, data, 0);
  }

The code in this except is executed by the "master" collector, e.g.,
the node with rank 2/3 `world.size()` in the `world` communicator and
rank 0 in the `local` (collector) communicator. It receives a message
from a generator via the `world` communicator, then broadcasts the
message to each of the collectors via the `local` communicator.

For more control in the creation of communicators for subgroups of
processes, the Boost.MPI [classref boost::mpi::group `group`] provides
facilities to compute the union (`|`), intersection (`&`), and
difference (`-`) of two groups, generate arbitrary subgroups, etc.

[endsect]

[section:skeleton_and_content Separating structure from content]

When communicating data types over MPI that are not fundamental to MPI
(such as strings, lists, and user-defined data types), Boost.MPI must
first serialize these data types into a buffer and then communicate
them; the receiver then copies the results into a buffer before
deserializing into an object on the other end. For some data types,
this overhead can be eliminated by using [classref
boost::mpi::is_mpi_datatype `is_mpi_datatype`]. However,
variable-length data types such as strings and lists cannot be MPI
data types. 

Boost.MPI supports a second technique for improving performance by
separating the structure of these variable-length data structures from
the content stored in the data structures. This feature is only
beneficial when the shape of the data structure remains the same but
the content of the data structure will need to be communicated several
times. For instance, in a finite element analysis the structure of the
mesh may be fixed at the beginning of computation but the various
variables on the cells of the mesh (temperature, stress, etc.) will be
communicated many times within the iterative analysis process. In this
case, Boost.MPI allows one to first send the "skeleton" of the mesh
once, then transmit the "content" multiple times. Since the content
need not contain any information about the structure of the data type,
it can be transmitted without creating separate communication buffers.

To illustrate the use of skeletons and content, we will take a
somewhat more limited example wherein a master process generates
random number sequences into a list and transmits them to several
slave processes. The length of the list will be fixed at program
startup, so the content of the list (i.e., the current sequence of
numbers) can be transmitted efficiently. The complete example is
available in `example/random_content.cpp`. We being with the master
process (rank 0), which builds a list, communicates its structure via
a [funcref boost::mpi::skeleton `skeleton`], then repeatedly
generates random number sequences to be broadcast to the slave
processes via [classref boost::mpi::content `content`]:

  
    // Generate the list and broadcast its structure
    std::list<int> l(list_len);
    broadcast(world, mpi::skeleton(l), 0);

    // Generate content several times and broadcast out that content
    mpi::content c = mpi::get_content(l);
    for (int i = 0; i < iterations; ++i) {
      // Generate new random values
      std::generate(l.begin(), l.end(), &random);

      // Broadcast the new content of l
      broadcast(world, c, 0);
    }

    // Notify the slaves that we're done by sending all zeroes
    std::fill(l.begin(), l.end(), 0);
    broadcast(world, c, 0);


The slave processes have a very similar structure to the master. They
receive (via the [funcref boost::mpi::broadcast
`broadcast()`] call) the skeleton of the data structure, then use it
to build their own lists of integers. In each iteration, they receive
via another `broadcast()` the new content in the data structure and
compute some property of the data:


    // Receive the content and build up our own list
    std::list<int> l;
    broadcast(world, mpi::skeleton(l), 0);

    mpi::content c = mpi::get_content(l);
    int i = 0;
    do {
      broadcast(world, c, 0);

      if (std::find_if
           (l.begin(), l.end(),
            std::bind1st(std::not_equal_to<int>(), 0)) == l.end())
        break;

      // Compute some property of the data.

      ++i;
    } while (true);


The skeletons and content of any Serializable data type can be
transmitted either via the [memberref
boost::mpi::communicator::send `send`] and [memberref
boost::mpi::communicator::recv `recv`] members of the
[classref boost::mpi::communicator `communicator`] class
(for point-to-point communicators) or broadcast via the [funcref
boost::mpi::broadcast `broadcast()`] collective. When
separating a data structure into a skeleton and content, be careful
not to modify the data structure (either on the sender side or the
receiver side) without transmitting the skeleton again. Boost.MPI can
not detect these accidental modifications to the data structure, which
will likely result in incorrect data being transmitted or unstable
programs. 

[endsect]


[section:performance_optimizations Performance optimizations]
[section:serialization_optimizations Serialization optimizations]

To obtain optimal performance for small fixed-length data types not containing
any pointers it is very important to mark them using the type traits of
Boost.MPI and Boost.Serialization. 

It was alredy discussed that fixed length types containing no pointers can be 
using as [classref
boost::mpi::is_mpi_datatype `is_mpi_datatype`], e.g.:

  namespace boost { namespace mpi {
    template <>
    struct is_mpi_datatype<gps_position> : mpl::true_ { };
  } }

or the equivalent macro 

  BOOST_IS_MPI_DATATYPE(gps_position)
  
In addition it can give a substantial performance gain to turn off tracking
and versioning for these types, if no pointers to these types are used, by
using the traits classes or helper macros of Boost.Serialization:

  BOOST_CLASS_TRACKING(gps_position,track_never)
  BOOST_CLASS_IMPLEMENTATION(gps_position,object_serializable)

[endsect]
  
[section:homogeneous_machines Homogeneous machines]

More optimizations are possible on homogeneous machines, by avoiding
MPI_Pack/MPI_Unpack calls but using direct bitwise copy. This feature can be
enabled by defining the macro BOOST_MPI_HOMOGENEOUS when building Boost.MPI and
when building the application.

In addition all classes need to be marked both as is_mpi_datatype and
as is_bitwise_serializable, by using the helper macro of Boost.Serialization:

  BOOST_IS_BITWISE_SERIALIZABLE(gps_position)

Usually it is safe to serialize a class for which is_mpi_datatype is true 
by using binary copy of the bits. The exception are classes for which
some members should be skipped for serialization.

[endsect]
[endsect]


[section:c_mapping Mapping from C MPI to Boost.MPI]

This section provides tables that map from the functions and constants
of the standard C MPI to their Boost.MPI equivalents. It will be most
useful for users that are already familiar with the C or Fortran
interfaces to MPI, or for porting existing parallel programs to Boost.MPI.

[table Point-to-point communication
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[`MPI_ANY_SOURCE`] [`any_source`]]

  [[`MPI_ANY_TAG`] [`any_tag`]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node40.html#Node40
`MPI_Bsend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Bsend_init`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node42.html#Node42
`MPI_Buffer_attach`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node42.html#Node42
`MPI_Buffer_detach`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node50.html#Node50
`MPI_Cancel`]] 
   [[memberref boost::mpi::request::cancel
`request::cancel`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node35.html#Node35
`MPI_Get_count`]] 
   [[memberref boost::mpi::status::count `status::count`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node46.html#Node46
`MPI_Ibsend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node50.html#Node50
`MPI_Iprobe`]]
   [[memberref boost::mpi::communicator::iprobe `communicator::iprobe`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node46.html#Node46
`MPI_Irsend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node46.html#Node46
`MPI_Isend`]] 
   [[memberref boost::mpi::communicator::isend
`communicator::isend`]]] 

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node46.html#Node46
`MPI_Issend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node46.html#Node46
`MPI_Irecv`]] 
   [[memberref boost::mpi::communicator::isend
`communicator::irecv`]]] 

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node50.html#Node50
`MPI_Probe`]]
   [[memberref boost::mpi::communicator::probe `communicator::probe`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node53.html#Node53
`MPI_PROC_NULL`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node34.html#Node34 `MPI_Recv`]]
   [[memberref boost::mpi::communicator::recv
`communicator::recv`]]] 

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Recv_init`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Request_free`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node40.html#Node40
`MPI_Rsend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Rsend_init`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node31.html#Node31
`MPI_Send`]]
   [[memberref boost::mpi::communicator::send
`communicator::send`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node52.html#Node52
`MPI_Sendrecv`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node52.html#Node52
`MPI_Sendrecv_replace`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Send_init`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node40.html#Node40
`MPI_Ssend`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Ssend_init`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Start`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node51.html#Node51
`MPI_Startall`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Test`]] [[memberref boost::mpi::request::wait `request::test`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Testall`]] [[funcref boost::mpi::test_all `test_all`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Testany`]] [[funcref boost::mpi::test_any `test_any`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Testsome`]] [[funcref boost::mpi::test_some `test_some`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node50.html#Node50
`MPI_Test_cancelled`]] 
   [[memberref boost::mpi::status::cancelled
`status::cancelled`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Wait`]] [[memberref boost::mpi::request::wait
`request::wait`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Waitall`]] [[funcref boost::mpi::wait_all `wait_all`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Waitany`]] [[funcref boost::mpi::wait_any `wait_any`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node47.html#Node47
`MPI_Waitsome`]] [[funcref boost::mpi::wait_some `wait_some`]]]
]

Boost.MPI automatically maps C and C++ data types to their MPI
equivalents. The following table illustrates the mappings between C++
types and MPI datatype constants.

[table Datatypes
  [[C Constant] [Boost.MPI Equivalent]]

  [[`MPI_CHAR`] [`signed char`]]
  [[`MPI_SHORT`] [`signed short int`]]
  [[`MPI_INT`] [`signed int`]]
  [[`MPI_LONG`] [`signed long int`]]
  [[`MPI_UNSIGNED_CHAR`] [`unsigned char`]]
  [[`MPI_UNSIGNED_SHORT`] [`unsigned short int`]]
  [[`MPI_UNSIGNED_INT`] [`unsigned int`]]
  [[`MPI_UNSIGNED_LONG`] [`unsigned long int`]]
  [[`MPI_FLOAT`] [`float`]]
  [[`MPI_DOUBLE`] [`double`]]
  [[`MPI_LONG_DOUBLE`] [`long double`]]
  [[`MPI_BYTE`] [unused]]
  [[`MPI_PACKED`] [used internally for [link
mpi.user_data_types serialized data types]]]
  [[`MPI_LONG_LONG_INT`] [`long long int`, if supported by compiler]]
  [[`MPI_UNSIGNED_LONG_LONG_INT`] [`unsigned long long int`, if
supported by compiler]]
  [[`MPI_FLOAT_INT`] [`std::pair<float, int>`]]
  [[`MPI_DOUBLE_INT`] [`std::pair<double, int>`]]
  [[`MPI_LONG_INT`] [`std::pair<long, int>`]]
  [[`MPI_2INT`] [`std::pair<int, int>`]]
  [[`MPI_SHORT_INT`] [`std::pair<short, int>`]]
  [[`MPI_LONG_DOUBLE_INT`] [`std::pair<long double, int>`]]
]

Boost.MPI does not provide direct wrappers to the MPI derived
datatypes functionality. Instead, Boost.MPI relies on the
_Serialization_ library to construct MPI datatypes for user-defined
classe. The section on [link mpi.user_data_types user-defined
data types] describes this mechanism, which is used for types that
marked as "MPI datatypes" using [classref
boost::mpi::is_mpi_datatype `is_mpi_datatype`].

The derived datatypes table that follows describes which C++ types
correspond to the functionality of the C MPI's datatype
constructor. Boost.MPI may not actually use the C MPI function listed
when building datatypes of a certain form. Since the actual datatypes
built by Boost.MPI are typically hidden from the user, many of these
operations are called internally by Boost.MPI.

[table Derived datatypes
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node56.html#Node56
`MPI_Address`]] [used automatically in Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node58.html#Node58
`MPI_Type_commit`]] [used automatically in Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_contiguous`]] [arrays]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node56.html#Node56
`MPI_Type_extent`]] [used automatically in Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node58.html#Node58
`MPI_Type_free`]] [used automatically in Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_hindexed`]] [any type used as a subobject]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_hvector`]] [unused]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_indexed`]] [any type used as a subobject]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node57.html#Node57
`MPI_Type_lb`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node56.html#Node56
`MPI_Type_size`]] [used automatically in Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_struct`]] [user-defined classes and structs]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node57.html#Node57
`MPI_Type_ub`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node55.html#Node55
`MPI_Type_vector`]] [used automatically in Boost.MPI]]
]

MPI's packing facilities store values into a contiguous buffer, which
can later be transmitted via MPI and unpacked into separate values via
MPI's unpacking facilities. As with datatypes, Boost.MPI provides an
abstract interface to MPI's packing and unpacking facilities. In
particular, the two archive classes [classref
boost::mpi::packed_oarchive `packed_oarchive`] and [classref
boost::mpi::packed_iarchive `packed_iarchive`] can be used
to pack or unpack a contiguous buffer using MPI's facilities.

[table Packing and unpacking
  [[C Function] [Boost.MPI Equivalent]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node62.html#Node62
`MPI_Pack`]] [[classref
boost::mpi::packed_oarchive `packed_oarchive`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node62.html#Node62
`MPI_Pack_size`]] [used internally by Boost.MPI]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node62.html#Node62
`MPI_Unpack`]] [[classref
boost::mpi::packed_iarchive `packed_iarchive`]]]
]

Boost.MPI supports a one-to-one mapping for most of the MPI
collectives. For each collective provided by Boost.MPI, the underlying
C MPI collective will be invoked when it is possible (and efficient)
to do so.

[table Collectives
  [[C Function] [Boost.MPI Equivalent]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node73.html#Node73
`MPI_Allgather`]] [[funcref boost::mpi::all_gather `all_gather`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node73.html#Node73
`MPI_Allgatherv`]] [most uses supported by [funcref boost::mpi::all_gather `all_gather`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node82.html#Node82
`MPI_Allreduce`]] [[funcref boost::mpi::all_reduce `all_reduce`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node75.html#Node75
`MPI_Alltoall`]] [[funcref boost::mpi::all_to_all `all_to_all`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node75.html#Node75
`MPI_Alltoallv`]] [most uses supported by [funcref boost::mpi::all_to_all `all_to_all`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node66.html#Node66
`MPI_Barrier`]] [[memberref
boost::mpi::communicator::barrier `communicator::barrier`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node67.html#Node67
`MPI_Bcast`]] [[funcref boost::mpi::broadcast `broadcast`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node69.html#Node69
`MPI_Gather`]] [[funcref boost::mpi::gather `gather`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node69.html#Node69
`MPI_Gatherv`]] [most uses supported by [funcref boost::mpi::gather `gather`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node77.html#Node77
`MPI_Reduce`]] [[funcref boost::mpi::reduce `reduce`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node83.html#Node83
`MPI_Reduce_scatter`]] [unsupported]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node84.html#Node84
`MPI_Scan`]] [[funcref boost::mpi::scan `scan`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node71.html#Node71
`MPI_Scatter`]] [[funcref boost::mpi::scatter `scatter`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node71.html#Node71
`MPI_Scatterv`]] [most uses supported by [funcref boost::mpi::scatter `scatter`]]]
]

Boost.MPI uses function objects to specify how reductions should occur
in its equivalents to `MPI_Allreduce`, `MPI_Reduce`, and
`MPI_Scan`. The following table illustrates how
[@http://www.mpi-forum.org/docs/mpi-11-html/node78.html#Node78
predefined] and
[@http://www.mpi-forum.org/docs/mpi-11-html/node80.html#Node80
user-defined] reduction operations can be mapped between the C MPI and
Boost.MPI.

[table Reduction operations
  [[C Constant] [Boost.MPI Equivalent]]

  [[`MPI_BAND`] [[classref boost::mpi::bitwise_and `bitwise_and`]]]
  [[`MPI_BOR`] [[classref boost::mpi::bitwise_or `bitwise_or`]]]
  [[`MPI_BXOR`] [[classref boost::mpi::bitwise_xor `bitwise_xor`]]]
  [[`MPI_LAND`] [`std::logical_and`]]
  [[`MPI_LOR`] [`std::logical_or`]]
  [[`MPI_LXOR`] [[classref boost::mpi::logical_xor `logical_xor`]]]
  [[`MPI_MAX`] [[classref boost::mpi::maximum `maximum`]]]
  [[`MPI_MAXLOC`] [unsupported]]
  [[`MPI_MIN`] [[classref boost::mpi::minimum `minimum`]]]
  [[`MPI_MINLOC`] [unsupported]]
  [[`MPI_Op_create`] [used internally by Boost.MPI]]
  [[`MPI_Op_free`] [used internally by Boost.MPI]]
  [[`MPI_PROD`] [`std::multiplies`]]
  [[`MPI_SUM`] [`std::plus`]]
]

MPI defines several special communicators, including `MPI_COMM_WORLD`
(including all processes that the local process can communicate with),
`MPI_COMM_SELF` (including only the local process), and
`MPI_COMM_EMPTY` (including no processes). These special communicators
are all instances of the [classref boost::mpi::communicator
`communicator`] class in Boost.MPI.

[table Predefined communicators
  [[C Constant] [Boost.MPI Equivalent]]

  [[`MPI_COMM_WORLD`] [a default-constructed [classref boost::mpi::communicator `communicator`]]]
  [[`MPI_COMM_SELF`] [a [classref boost::mpi::communicator `communicator`] that contains only the current process]]
  [[`MPI_COMM_EMPTY`] [a [classref boost::mpi::communicator `communicator`] that evaluates false]]
]

Boost.MPI supports groups of processes through its [classref
boost::mpi::group `group`] class.

[table Group operations and constants
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[`MPI_GROUP_EMPTY`] [a default-constructed [classref
  boost::mpi::group `group`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node97.html#Node97
  `MPI_Group_size`]] [[memberref boost::mpi::group::size `group::size`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node97.html#Node97
  `MPI_Group_rank`]] [memberref boost::mpi::group::rank `group::rank`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node97.html#Node97
  `MPI_Group_translate_ranks`]] [memberref boost::mpi::group::translate_ranks `group::translate_ranks`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node97.html#Node97
  `MPI_Group_compare`]] [operators `==` and `!=`]]
  [[`MPI_IDENT`] [operators `==` and `!=`]]
  [[`MPI_SIMILAR`] [operators `==` and `!=`]]
  [[`MPI_UNEQUAL`] [operators `==` and `!=`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Comm_group`]] [[memberref
  boost::mpi::communicator::group `communicator::group`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_union`]] [operator `|` for groups]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_intersection`]] [operator `&` for groups]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_difference`]] [operator `-` for groups]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_incl`]] [[memberref boost::mpi::group::include `group::include`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_excl`]] [[memberref boost::mpi::group::include `group::exclude`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_range_incl`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node98.html#Node98
  `MPI_Group_range_excl`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node99.html#Node99
  `MPI_Group_free`]] [used automatically in Boost.MPI]]
]

Boost.MPI provides manipulation of communicators through the [classref
boost::mpi::communicator `communicator`] class.

[table Communicator operations
  [[C Function] [Boost.MPI Equivalent]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node101.html#Node101
  `MPI_Comm_size`]] [[memberref boost::mpi::communicator::size `communicator::size`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node101.html#Node101
  `MPI_Comm_rank`]] [[memberref boost::mpi::communicator::rank
  `communicator::rank`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node101.html#Node101
  `MPI_Comm_compare`]] [operators `==` and `!=`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node102.html#Node102
  `MPI_Comm_dup`]] [[classref boost::mpi::communicator `communicator`]
  class constructor using `comm_duplicate`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node102.html#Node102
  `MPI_Comm_create`]] [[classref boost::mpi::communicator
  `communicator`] constructor]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node102.html#Node102
  `MPI_Comm_split`]] [[memberref boost::mpi::communicator::split
  `communicator::split`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node103.html#Node103
  `MPI_Comm_free`]] [used automatically in Boost.MPI]]
]

Boost.MPI currently provides support for inter-communicators via the
[classref boost::mpi::intercommunicator `intercommunicator`] class.

[table Inter-communicator operations
  [[C Function] [Boost.MPI Equivalent]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node112.html#Node112
  `MPI_Comm_test_inter`]] [use [memberref boost::mpi::communicator::as_intercommunicator `communicator::as_intercommunicator`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node112.html#Node112
  `MPI_Comm_remote_size`]] [[memberref boost::mpi::intercommunicator::remote_size] `intercommunicator::remote_size`]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node112.html#Node112
  `MPI_Comm_remote_group`]] [[memberref boost::mpi::intercommunicator::remote_group `intercommunicator::remote_group`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node113.html#Node113
  `MPI_Intercomm_create`]] [[classref boost::mpi::intercommunicator `intercommunicator`] constructor]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node113.html#Node113
  `MPI_Intercomm_merge`]] [[memberref boost::mpi::intercommunicator::merge `intercommunicator::merge`]]]
]

Boost.MPI currently provides no support for attribute caching.

[table Attributes and caching
 [[C Function/Constant] [Boost.MPI Equivalent]]

 [[`MPI_NULL_COPY_FN`] [unsupported]]
 [[`MPI_NULL_DELETE_FN`] [unsupported]]
 [[`MPI_KEYVAL_INVALID`] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Keyval_create`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Copy_function`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Delete_function`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Keyval_free`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Attr_put`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Attr_get`]] [unsupported]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node119.html#Node119
 `MPI_Attr_delete`]] [unsupported]]
]

Boost.MPI will provide complete support for creating communicators
with different topologies and later querying those topologies. Support
for graph topologies is provided via an interface to the
[@http://www.boost.org/libs/graph/doc/index.html Boost Graph Library
(BGL)], where a communicator can be created which matches the
structure of any BGL graph, and the graph topology of a communicator
can be viewed as a BGL graph for use in existing, generic graph
algorithms.

[table Process topologies
  [[C Function/Constant] [Boost.MPI Equivalent]]
  
  [[`MPI_GRAPH`] [unnecessary; use [memberref boost::mpi::communicator::has_graph_topology `communicator::has_graph_topology`]]]
  [[`MPI_CART`] [unnecessary; use [memberref boost::mpi::communicator::has_cartesian_topology `communicator::has_cartesian_topology`]]]

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node133.html#Node133
  `MPI_Cart_create`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node134.html#Node134
  `MPI_Dims_create`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node135.html#Node135
  `MPI_Graph_create`]] [[memberref
  boost::mpi::communicator::with_graph_topology
  `communicator::with_graph_topology`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Topo_test`]] [[memberref
  boost::mpi::communicator::has_graph_topology
  `communicator::has_graph_topology`], [memberref
  boost::mpi::communicator::has_cartesian_topology
  `communicator::has_cartesian_topology`]]] 
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Graphdims_get`]] [[funcref boost::mpi::num_vertices
  `num_vertices`], [funcref boost::mpi::num_edges `num_edges`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Graph_get`]] [[funcref boost::mpi::vertices
  `vertices`], [funcref boost::mpi::edges `edges`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Cartdim_get`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Cart_get`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Cart_rank`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Cart_coords`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Graph_neighbors_count`]] [[funcref boost::mpi::out_degree
  `out_degree`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node136.html#Node136
  `MPI_Graph_neighbors`]] [[funcref boost::mpi::out_edges
  `out_edges`], [funcref boost::mpi::adjacent_vertices `adjacent_vertices`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node137.html#Node137
  `MPI_Cart_shift`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node138.html#Node138
  `MPI_Cart_sub`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node139.html#Node139
  `MPI_Cart_map`]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node139.html#Node139
  `MPI_Graph_map`]] [unsupported]]
]

Boost.MPI supports environmental inquires through the [classref
boost::mpi::environment `environment`] class.

[table Environmental inquiries
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[`MPI_TAG_UB`] [unnecessary; use [memberref
  boost::mpi::environment::max_tag `environment::max_tag`]]]
  [[`MPI_HOST`] [unnecessary; use [memberref
  boost::mpi::environment::host_rank `environment::host_rank`]]]
  [[`MPI_IO`] [unnecessary; use [memberref
  boost::mpi::environment::io_rank `environment::io_rank`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node143.html#Node147
  `MPI_Get_processor_name`]] 
  [[memberref boost::mpi::environment::processor_name
  `environment::processor_name`]]]
]

Boost.MPI translates MPI errors into exceptions, reported via the
[classref boost::mpi::exception `exception`] class.

[table Error handling
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[`MPI_ERRORS_ARE_FATAL`] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[`MPI_ERRORS_RETURN`] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
  `MPI_errhandler_create`]] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
  `MPI_errhandler_set`]] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
  `MPI_errhandler_get`]] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
  `MPI_errhandler_free`]] [unused; errors are translated into
  Boost.MPI exceptions]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
  `MPI_Error_string`]] [used internally by Boost.MPI]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node149.html#Node149
  `MPI_Error_class`]] [[memberref boost::mpi::exception::error_class `exception::error_class`]]]
]

The MPI timing facilities are exposed via the Boost.MPI [classref
boost::mpi::timer `timer`] class, which provides an interface
compatible with the [@http://www.boost.org/libs/timer/index.html Boost
Timer library].

[table Timing facilities
  [[C Function/Constant] [Boost.MPI Equivalent]]

  [[`MPI_WTIME_IS_GLOBAL`] [unnecessary; use [memberref
  boost::mpi::timer::time_is_global `timer::time_is_global`]]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node150.html#Node150
  `MPI_Wtime`]] [use [memberref boost::mpi::timer::elapsed
  `timer::elapsed`] to determine the time elapsed from some specific
  starting point]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node150.html#Node150
  `MPI_Wtick`]] [[memberref boost::mpi::timer::elapsed_min `timer::elapsed_min`]]]
]

MPI startup and shutdown are managed by the construction and
descruction of the Boost.MPI [classref boost::mpi::environment
`environment`] class.

[table Startup/shutdown facilities
  [[C Function] [Boost.MPI Equivalent]]       

  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node151.html#Node151
  `MPI_Init`]] [[classref boost::mpi::environment `environment`]
  constructor]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node151.html#Node151
  `MPI_Finalize`]] [[classref boost::mpi::environment `environment`]
  destructor]]
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node151.html#Node151
  `MPI_Initialized`]] [[memberref boost::mpi::environment::initialized
  `environment::initialized`]]] 
 [[[@http://www.mpi-forum.org/docs/mpi-11-html/node151.html#Node151
  `MPI_Abort`]] [[memberref boost::mpi::environment::abort
  `environment::abort`]]] 
]

Boost.MPI does not provide any support for the profiling facilities in
MPI 1.1. 

[table Profiling interface
  [[C Function] [Boost.MPI Equivalent]]
  
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node153.html#Node153
  `PMPI_*` routines]] [unsupported]]
  [[[@http://www.mpi-forum.org/docs/mpi-11-html/node156.html#Node156
  `MPI_Pcontrol`]] [unsupported]]
]

[endsect]

[endsect]

[xinclude mpi_autodoc.xml]

[section:python Python Bindings]
[python]

Boost.MPI provides an alternative MPI interface from the _Python_
programming language via the `boost.mpi` module. The
Boost.MPI Python bindings, built on top of the C++ Boost.MPI using the
_BoostPython_ library, provide nearly all of the functionality of
Boost.MPI within a dynamic, object-oriented language.

The Boost.MPI Python module can be built and installed from the
`libs/mpi/build` directory. Just follow the [link
mpi.config configuration] and [link mpi.installation
installation] instructions for the C++ Boost.MPI. Once you have
installed the Python module, be sure that the installation location is
in your `PYTHONPATH`.

[section:python_quickstart Quickstart]

[python]

Getting started with the Boost.MPI Python module is as easy as
importing `boost.mpi`. Our first "Hello, World!" program is
just two lines long:

  import boost.mpi as mpi
  print "I am process %d of %d." % (mpi.rank, mpi.size)

Go ahead and run this program with several processes. Be sure to
invoke the `python` interpreter from `mpirun`, e.g., 
  
[pre
mpirun -np 5 python hello_world.py
]

This will return output such as:

[pre
I am process 1 of 5.
I am process 3 of 5.
I am process 2 of 5.
I am process 4 of 5.
I am process 0 of 5.
]

Point-to-point operations in Boost.MPI have nearly the same syntax in
Python as in C++. We can write a simple two-process Python program
that prints "Hello, world!" by transmitting Python strings:

  import boost.mpi as mpi

  if mpi.world.rank == 0:
    mpi.world.send(1, 0, 'Hello')
    msg = mpi.world.recv(1, 1)
    print msg,'!'
  else:
    msg = mpi.world.recv(0, 0)
    print (msg + ', '),
    mpi.world.send(0, 1, 'world')

There are only a few notable differences between this Python code and
the example [link mpi.point_to_point in the C++
tutorial]. First of all, we don't need to write any initialization
code in Python: just loading the `boost.mpi` module makes the
appropriate `MPI_Init` and `MPI_Finalize` calls. Second, we're passing
Python objects from one process to another through MPI. Any Python
object that can be pickled can be transmitted; the next section will
describe in more detail how the Boost.MPI Python layer transmits
objects. Finally, when we receive objects with `recv`, we don't need
to specify the type because transmission of Python objects is
polymorphic. 

When experimenting with Boost.MPI in Python, don't forget that help is
always available via `pydoc`: just pass the name of the module or
module entity on the command line (e.g., `pydoc
boost.mpi.communicator`) to receive complete reference
documentation. When in doubt, try it!
[endsect]

[section:python_user_data Transmitting User-Defined Data]
Boost.MPI can transmit user-defined data in several different ways.
Most importantly, it can transmit arbitrary _Python_ objects by pickling
them at the sender and unpickling them at the receiver, allowing
arbitrarily complex Python data structures to interoperate with MPI.

Boost.MPI also supports efficient serialization and transmission of
C++ objects (that have been exposed to Python) through its C++
interface. Any C++ type that provides (de-)serialization routines that
meet the requirements of the Boost.Serialization library is eligible
for this optimization, but the type must be registered in advance. To
register a C++ type, invoke the C++ function [funcref
boost::mpi::python::register_serialized
register_serialized]. If your C++ types come from other Python modules
(they probably will!), those modules will need to link against the
`boost_mpi` and `boost_mpi_python` libraries as described in the [link
mpi.installation installation section]. Note that you do
*not* need to link against the Boost.MPI Python extension module.

Finally, Boost.MPI supports separation of the structure of an object
from the data it stores, allowing the two pieces to be transmitted
separately. This "skeleton/content" mechanism, described in more
detail in a later section, is a communication optimization suitable
for problems with fixed data structures whose internal data changes
frequently.
[endsect]

[section:python_collectives Collectives]

Boost.MPI supports all of the MPI collectives (`scatter`, `reduce`,
`scan`, `broadcast`, etc.) for any type of data that can be
transmitted with the point-to-point communication operations. For the
MPI collectives that require a user-specified operation (e.g., `reduce`
and `scan`), the operation can be an arbitrary Python function. For
instance, one could concatenate strings with `all_reduce`:

  mpi.all_reduce(my_string, lambda x,y: x + y)

The following module-level functions implement MPI collectives:
  all_gather    Gather the values from all processes.
  all_reduce    Combine the results from all processes.
  all_to_all    Every process sends data to every other process.
  broadcast     Broadcast data from one process to all other processes.
  gather        Gather the values from all processes to the root.
  reduce        Combine the results from all processes to the root.
  scan          Prefix reduction of the values from all processes.
  scatter       Scatter the values stored at the root to all processes.
[endsect]

[section:python_skeleton_content Skeleton/Content Mechanism]
Boost.MPI provides a skeleton/content mechanism that allows the
transfer of large data structures to be split into two separate stages,
with the skeleton (or, "shape") of the data structure sent first and
the content (or, "data") of the data structure sent later, potentially
several times, so long as the structure has not changed since the
skeleton was transferred. The skeleton/content mechanism can improve
performance when the data structure is large and its shape is fixed,
because while the skeleton requires serialization (it has an unknown
size), the content transfer is fixed-size and can be done without
extra copies.

To use the skeleton/content mechanism from Python, you must first
register the type of your data structure with the skeleton/content
mechanism *from C++*. The registration function is [funcref
boost::mpi::python::register_skeleton_and_content
register_skeleton_and_content] and resides in the [headerref
boost/mpi/python.hpp <boost/mpi/python.hpp>] header.

Once you have registered your C++ data structures, you can extract
the skeleton for an instance of that data structure with `skeleton()`.
The resulting `skeleton_proxy` can be transmitted via the normal send
routine, e.g.,

  mpi.world.send(1, 0, skeleton(my_data_structure))

`skeleton_proxy` objects can be received on the other end via `recv()`,
which stores a newly-created instance of your data structure with the
same "shape" as the sender in its `"object` attribute:

  shape = mpi.world.recv(0, 0)
  my_data_structure = shape.object

Once the skeleton has been transmitted, the content (accessed via 
`get_content`) can be transmitted in much the same way. Note, however,
that the receiver also specifies `get_content(my_data_structure)` in its
call to receive:

  if mpi.rank == 0:
    mpi.world.send(1, 0, get_content(my_data_structure))
  else:
    mpi.world.recv(0, 0, get_content(my_data_structure))

Of course, this transmission of content can occur repeatedly, if the
values in the data structure--but not its shape--changes.

The skeleton/content mechanism is a structured way to exploit the
interaction between custom-built MPI datatypes and `MPI_BOTTOM`, to
eliminate extra buffer copies.
[endsect]

[section:python_compatbility C++/Python MPI Compatibility]
Boost.MPI is a C++ library whose facilities have been exposed to Python
via the Boost.Python library. Since the Boost.MPI Python bindings are
build directly on top of the C++ library, and nearly every feature of
C++ library is available in Python, hybrid C++/Python programs using
Boost.MPI can interact, e.g., sending a value from Python but receiving
that value in C++ (or vice versa). However, doing so requires some
care. Because Python objects are dynamically typed, Boost.MPI transfers
type information along with the serialized form of the object, so that
the object can be received even when its type is not known. This
mechanism differs from its C++ counterpart, where the static types of
transmitted values are always known.

The only way to communicate between the C++ and Python views on
Boost.MPI is to traffic entirely in Python objects. For Python, this
is the normal state of affairs, so nothing will change. For C++, this
means sending and receiving values of type `boost::python::object`,
from the _BoostPython_ library. For instance, say we want to transmit
an integer value from Python:

  comm.send(1, 0, 17)

In C++, we would receive that value into a Python object and then
`extract` an integer value:

[c++]

  boost::python::object value;
  comm.recv(0, 0, value);
  int int_value = boost::python::extract<int>(value);

In the future, Boost.MPI will be extended to allow improved
interoperability with the C++ Boost.MPI and the C MPI bindings.
[endsect]

[section:pythonref Reference]
The Boost.MPI Python module, `boost.mpi`, has its own
[@boost.mpi.html reference documentation], which is also
available using `pydoc` (from the command line) or
`help(boost.mpi)` (from the Python interpreter).

[endsect]

[endsect]

[section:design Design Philosophy]

The design philosophy of the Parallel MPI library is very simple: be
both convenient and efficient. MPI is a library built for
high-performance applications, but it's FORTRAN-centric,
performance-minded design makes it rather inflexible from the C++
point of view: passing a string from one process to another is
inconvenient, requiring several messages and explicit buffering;
passing a container of strings from one process to another requires
an extra level of manual bookkeeping; and passing a map from strings
to containers of strings is positively infuriating. The Parallel MPI
library allows all of these data types to be passed using the same
simple `send()` and `recv()` primitives. Likewise, collective
operations such as [funcref boost::mpi::reduce `reduce()`]
allow arbitrary data types and function objects, much like the C++
Standard Library would. 

The higher-level abstractions provided for convenience must not have
an impact on the performance of the application. For instance, sending
an integer via `send` must be as efficient as a call to `MPI_Send`,
which means that it must be implemented by a simple call to
`MPI_Send`; likewise, an integer [funcref boost::mpi::reduce
`reduce()`] using `std::plus<int>` must be implemented with a call to
`MPI_Reduce` on integers using the `MPI_SUM` operation: anything less
will impact performance. In essence, this is the "don't pay for what
you don't use" principle: if the user is not transmitting strings,
s/he should not pay the overhead associated with strings. 

Sometimes, achieving maximal performance means foregoing convenient
abstractions and implementing certain functionality using lower-level
primitives. For this reason, it is always possible to extract enough
information from the abstractions in Boost.MPI to minimize
the amount of effort required to interface between Boost.MPI
and the C MPI library.
[endsect]

[section:performance Performance Evaluation]

Message-passing performance is crucial in high-performance distributed
computing. To evaluate the performance of Boost.MPI, we modified the
standard [@http://www.scl.ameslab.gov/netpipe/ NetPIPE] benchmark
(version 3.6.2) to use Boost.MPI and compared its performance against
raw MPI. We ran five different variants of the NetPIPE benchmark:

# MPI: The unmodified NetPIPE benchmark.

# Boost.MPI: NetPIPE modified to use Boost.MPI calls for
  communication.

# MPI (Datatypes): NetPIPE modified to use a derived datatype (which
  itself contains a single `MPI_BYTE`) rathan than a fundamental
  datatype.

# Boost.MPI (Datatypes): NetPIPE modified to use a user-defined type
  `Char` in place of the fundamental `char` type. The `Char` type
  contains a single `char`, a `serialize()` method to make it
  serializable, and specializes [classref
  boost::mpi::is_mpi_datatype is_mpi_datatype] to force
  Boost.MPI to build a derived MPI data type for it.

# Boost.MPI (Serialized): NetPIPE modified to use a user-defined type
  `Char` in place of the fundamental `char` type. This `Char` type
  contains a single `char` and is serializable. Unlike the Datatypes
  case, [classref boost::mpi::is_mpi_datatype
  is_mpi_datatype] is *not* specialized, forcing Boost.MPI to perform
  many, many serialization calls.

The actual tests were performed on the Odin cluster in the
[@http://www.cs.indiana.edu/ Department of Computer Science] at
[@http://www.iub.edu Indiana University], which contains 128 nodes
connected via Infiniband. Each node contains 4GB memory and two AMD
Opteron processors. The NetPIPE benchmarks were compiled with Intel's
C++ Compiler, version 9.0, Boost 1.35.0 (prerelease), and
[@http://www.open-mpi.org/ Open MPI] version 1.1. The NetPIPE results
follow:

[$../../libs/mpi/doc/netpipe.png]

There are a some observations we can make about these NetPIPE
results. First of all, the top two plots show that Boost.MPI performs
on par with MPI for fundamental types. The next two plots show that
Boost.MPI performs on par with MPI for derived data types, even though
Boost.MPI provides a much more abstract, completely transparent
approach to building derived data types than raw MPI. Overall
performance for derived data types is significantly worse than for
fundamental data types, but the bottleneck is in the underlying MPI
implementation itself. Finally, when forcing Boost.MPI to serialize
characters individually, performance suffers greatly. This particular
instance is the worst possible case for Boost.MPI, because we are
serializing millions of individual characters.  Overall, the
additional abstraction provided by Boost.MPI does not impair its
performance.

[endsect]

[section:history Revision History]

* *Boost 1.36.0*: 
  * Support for non-blocking operations in Python, from Andreas Klöckner

* *Boost 1.35.0*: Initial release, containing the following post-review changes 
  * Support for arrays in all collective operations
  * Support default-construction of [classref boost::mpi::environment environment]
    
* *2006-09-21*: Boost.MPI accepted into Boost.

[endsect]

[section:acknowledge Acknowledgments]
Boost.MPI was developed with support from Zurcher Kantonalbank. Daniel
Egloff and Michael Gauckler contributed many ideas to Boost.MPI's
design, particularly in the design of its abstractions for
MPI data types and the novel skeleton/context mechanism for large data
structures. Prabhanjan (Anju) Kambadur developed the predecessor to
Boost.MPI that proved the usefulness of the Serialization library in
an MPI setting and the performance benefits of specialization in a C++
abstraction layer for MPI. Jeremy Siek managed the formal review of Boost.MPI.

[endsect]