MPI

A Primer on Parallel Programming

To pull a bigger wagon it is easier to add more oxen than to find (or build) a bigger ox.

—Gropp, Lusk & Skjellum, Using MPI

To build a house as quickly as possible, we do not look for the fastest person to do all the work but instead we hire as many people as required and spread the work among them so that various construction tasks are performed at the same time --- "in parallel". Computational problems are conceptually similar. Since there is a limit to how fast a single machine can perform, we attempt to divide up the computational problem at hand and assign work to be completed in parallel to multiple computers. This approach is important not only in speeding up computations but also in tackling problems requiring large amounts of memory.

The most significant concept to master in designing and building parallel applications is communication. Complexity arises due to communication requirements. In order for multiple workers to accomplish a task in parallel, they need to be able to communicate with one another. In the context of software, we have many processes each working on part of a solution, needing values that were computed ---or are yet to be computed!--- by other processes.

There are two major models of computational parallelism: shared memory, and distributed memory.

In shared memory parallelism (commonly and casually abbreviated SMP), all processors see the same memory image, or to put it another way, all memory is globally addressable and all the processes can ultimately access it. Communication between processes on an SMP machine is implicit --- any process can ultimately read and write values to memory that can be subsequently accessed and manipulated directly by others. The challenge in writing these kinds of programs is data consistency: one should take extra care to ensure data is not modified by more than one process at a time.

Figure 1: A conceptual picture of a shared memory architecture

Distributed memory parallelism is equivalent to a collection of workstations linked by a dedicated network for communication: a cluster. In this model, processes each have their own private memory, and may run on physically distinct machines. When processes need to communicate, they do so by sending messages. A process typically invokes a function to send data and the destination process invokes a function to receive it. A major challenge in distributed memory programming is how to minimize communication overhead. Networks, even the fastest dedicated hardware interconnects, transmit data orders of magnitude slower than within a single machine. Memory access times are typically measured in ones to hundreds of nanoseconds, while network latency is typically expressed in microseconds.

Figure 2: A conceptual picture of a cluster architecture

The remainder of this tutorial will consider distributed memory programming on a cluster using the Message Passing Interface.

What is MPI?

The Message Passing Interface (MPI) is, strictly speaking, a standard describing a set of subroutines, functions, objects, etc., with which one can write parallel programs in a distributed memory environment. Many different implementations of the standard have been produced, such as Open MPI, Intel MPI, MPICH, and MVAPICH. The standard describes how MPI should be called from Fortran, C, and C++ languages, but unofficial "bindings" can be found for several other languages. Note that MPI 3.0 dropped official C++ bindings but instead you can use the C bindings from C++, or Boost MPI. For Python we give examples using the MPI for Python package MPI4py.

Since MPI is an open, non-proprietary standard, an MPI program can easily be ported to many different computers. Applications that use it can run on a large number of cores at once, often with good parallel efficiency (see the scalability page for more details). Given that memory is local to each process, some aspects of debugging are simplified --- it isn't possible for one process to interfere with the memory of another, and if a program generates a segmentation fault the resulting core file can be processed by standard serial debugging tools. However, due to the need to manage communication and synchronization explicitly, MPI programs may appear more complex than programs written with tools that support implicit communication. Furthermore, in designing an MPI program one should take care to minimize communication overhead to achieve a good speed-up from the parallel computation.

In the following we will highlight a few of these issues and discuss strategies to deal with them. Suggested references are presented at the end of this tutorial and the reader is encouraged to consult them for additional information.

MPI Programming Basics

This tutorial will present the development of an MPI code in C, C++, Fortran, and Python, but the concepts apply to any language for which MPI bindings exist. For simplicity our goal will be to parallelize the venerable "Hello, World!" program, which appears below for reference.

CC++FortranPython

File : hello.c

#include <stdio.h>
 
 int main()
 {
     printf("Hello, world!\n");
 
     return(0);
 }

File : hello.cpp

#include <iostream>
using namespace std;

int main()
{
    cout << "Hello, world!" << endl;
    return 0;
}

File : hello.f90

program hello
 
     print *, 'Hello, world!'
 
 end program hello

File : hello.py

print('Hello, world!')

Compiling and running the program looks something like this:

[~]$ vi hello.c
[~]$ cc -Wall hello.c -o hello
[~]$ ./hello 
Hello, world!

SPMD Programming

Parallel programs written using MPI make use of an execution model called Single Program, Multiple Data, or SPMD. The SPMD model involves running a number of copies of a single program. In MPI, each copy or "process" is assigned a unique number, referred to as the rank of the process, and each process can obtain its rank when it runs. When a process should behave differently, we usually use an "if" statement based on the rank of the process to execute the appropriate set of instructions.

Figure 3: SPMD model illustrating conditional branching to control divergent behaviour

Framework

Each MPI program must include the relevant header file or use the relevant module (mpi.h for C/C++, mpif.h, use mpi, or use mpi_f08 for Fortran, where mpif.h is strongly discouraged and mpi_f08 recommended for new Fortran 2008 code), and be compiled and linked against the desired MPI implementation. Most MPI implementations provide a handy script, often called a compiler wrapper, that handles all set-up issues with respect to include and lib directories, linking flags, etc. Our examples will all use these compiler wrappers:

• C language wrapper: mpicc
• Fortran: mpifort (recommended) or mpif90
• C++: mpiCC or mpicxx

The copies of an MPI program, once they start running, must coordinate with one another somehow. This cooperation starts when each one calls an initialization function before it uses any other MPI features. The prototype for this function appears below:

CBoost (C++)FortranFortran 2008Python (mpi4py)

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

boost::mpi::environment(int &, char **&, bool = true);

 MPI_INIT(IERR)
 INTEGER :: IERR

 MPI_Init(ierr)
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

# importing automatically initializes MPI with mpi4py
MPI.Init()

The arguments to the C MPI_Init are pointers to the argc and argv variables that represent the command-line arguments to the program. Like all C MPI functions, the return value represents the error status of the function. Fortran MPI subroutines return the error status in an additional argument, IERR, which is optional if you use mpi_f08.

Similarly, we must call a function MPI_Finalize to do any clean-up that might be required before our program exits. The prototype for this function appears below:

CBoost (C++)FortranFortran 2008Python (mpi4py)

 int MPI_Finalize(void);

Nothing needed

 MPI_FINALIZE(IERR)
 INTEGER :: IERR

 MPI_Finalize(ierr)
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

# mpi4py installs a termination hook so there is no need to explicitly call MPI.Finalize.
MPI.Finalize()

As a rule of thumb, it is a good idea to call MPI_Init as the first statement of our program, and MPI_Finalize as its last statement. Let's now modify our "Hello, world!" program accordingly.

CBoost (C++)FortranFortran 2008Python (mpi4py)

File : phello0.c

#include <stdio.h>
#include <mpi.h>
 
int main(int argc, char *argv[])
{
    MPI_Init(&argc, &argv);
 
    printf("Hello, world!\n");
 
    MPI_Finalize();
    return(0);
}

File : phello0.cpp

#include <iostream>
#include <boost/mpi.hpp>
using namespace std;
using namespace boost;

int main(int argc, char *argv[])
{
    mpi::environment env(argc, argv);
    cout << "Hello, world!" << endl;
    return 0;
}

File : phello0.f90

program phello0
 
     use mpi
     implicit none
 
     integer :: ierror
 
     call MPI_INIT(ierror)
     print *, 'Hello, world!'
     call MPI_FINALIZE(ierror)

 end program phello0

File : phello0.f90

program phello0
 
     use mpi_f08
     implicit none
 
     call MPI_Init()
     print *, 'Hello, world!'
     call MPI_Finalize()

 end program phello0

File : phello0.py

from mpi4py import MPI
print('Hello, world!')

Rank and Size

We could now run this program under control of MPI, but each process would only output the original string which isn't very interesting. Let's instead have each process output its rank and how many processes are running in total. This information is obtained at run-time by the use of the following functions:

CBoost (C++)FortranFortran 2008Python (mpi4py)

 int MPI_Comm_size(MPI_Comm comm, int *nproc);
 int MPI_Comm_rank(MPI_Comm comm, int *myrank);

int mpi::communicator::size();
int mpi::communicator::rank();

 MPI_COMM_SIZE(COMM, NPROC, IERR)
 INTEGER :: COMM, NPROC, IERR
 
 MPI_COMM_RANK(COMM, RANK, IERR)
 INTEGER :: COMM, RANK, IERR

 MPI_Comm_size(comm, size, ierr)
 TYPE(MPI_Comm), INTENT(IN) :: comm
 INTEGER, INTENT(OUT) :: size
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

 MPI_Comm_rank(comm, rank, ierr)
 TYPE(MPI_Comm), INTENT(IN) :: comm
 INTEGER, INTENT(OUT) :: rank
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

 MPI.Intracomm.Get_rank(self)

 MPI.Intracomm.Get_size(self)

MPI_Comm_size reports the number of processes running as part of this job by assigning it to the result parameter nproc. Similarly, MPI_Comm_rank reports the rank of the calling process to the result parameter myrank. Ranks in MPI start from 0 rather than 1, so given N processes we expect the ranks to be 0..(N-1). The comm argument is a communicator, which is a set of processes capable of sending messages to one another. For the purpose of this tutorial we will always pass in the predefined value MPI_COMM_WORLD, which is simply all the MPI processes started with the job. It is possible to define and use your own communicators, but that is beyond the scope of this tutorial and the reader is referred to the provided references for additional detail.

Let us incorporate these functions into our program, and have each process output its rank and size information. Note that since all processes are still performing identical operations, there are no conditional blocks required in the code.

CBoost (C++)FortranFortran 2008Python (mpi4py)

File : phello1.c

#include <stdio.h>
 #include <mpi.h>
 
 int main(int argc, char *argv[])
 {
     int rank, size;
 
     MPI_Init(&argc, &argv);
     MPI_Comm_rank(MPI_COMM_WORLD, &rank);
     MPI_Comm_size(MPI_COMM_WORLD, &size);
 
     printf("Hello, world! "
             "from process %d of %d\n", rank, size);
 
     MPI_Finalize();
     return(0);
 }

File : phello1.cpp

#include <iostream>
#include <boost/mpi.hpp>
using namespace std;
using namespace boost;

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

    cout << "Hello, world! from process " << world.rank() << " of " << world.size() << endl;
    return 0;
}

File : phello1.f90

program phello1
 
    use mpi
    implicit none

    integer :: rank, size, ierror
 
    call MPI_INIT(ierror)
    call MPI_COMM_SIZE(MPI_COMM_WORLD, size, ierror)
    call MPI_COMM_RANK(MPI_COMM_WORLD, rank, ierror)
 
    print *, 'Hello from process ', rank, ' of ', size
 
    call MPI_FINALIZE(ierror)
 
 end program phello1

File : phello1.f90

program phello1
 
    use mpi_f08
    implicit none
 
    integer :: rank, size
 
    call MPI_Init()
    call MPI_Comm_size(MPI_COMM_WORLD, size)
    call MPI_Comm_rank(MPI_COMM_WORLD, rank)
 
    print *, 'Hello from process ', rank, ' of ', size
 
    call MPI_Finalize(ierror)
 
 end program phello1

File : phello1.py

from mpi4py import MPI
 
comm = MPI.COMM_WORLD
size = comm.Get_size()
rank = comm.Get_rank() 

print('Hello from process %d of %d'%(rank, size))

Compile and run this program using 2, 4 and 8 processes. Note that each running process produces output based on the values of its local variables. The stdout of all running processes is simply concatenated together. As you run the program using more processes, you may see that the output from the different processes does not appear in order or rank: You should make no assumptions about the order of output from different processes.

[~]$ vi phello1.c 
[~]$ mpicc -Wall phello1.c -o phello1
[~]$ mpirun -np 4 ./phello1
Hello, world! from process 0 of 4
Hello, world! from process 2 of 4
Hello, world! from process 1 of 4
Hello, world! from process 3 of 4

If you are using the Boost version, you should compile with:

[~]$ mpic++ --std=c++11 phello1.cpp -lboost_mpi-mt -lboost_serialization-mt -o phello1

If you are using the Python version, you don't need to compile but can run with:

[~]$ mpirun -np 4 python phello1.py

Communication

While we now have a parallel version of our "Hello, World!" program, it isn't very interesting as there is no communication between the processes. Let's fix this by having the processes send messages to one another.

We'll have each process send the string "hello" to the one with the next higher rank number. Rank i will send its message to rank i+1, and we'll have the last process, rank N-1, send its message back to process 0 (a nice communication ring!). A short way to express this is process i sends to process (i+1)%N, where there are N processes and % is the modulus operator.

MPI provides a large number of functions for sending and receiving data of almost any composition in a variety of communication patterns (one-to-one, one-to-many, many-to-one, and many-to-many). But the simplest functions to understand are the ones that send a sequence of one or more instances of an atomic data type from one process to another, MPI_Send and MPI_Recv.

A process sends data by calling the MPI_Send function. Referring to the following function prototypes, MPI_Send can be summarized as sending count contiguous instances of datatype to the process with the specified rank, and the data is in the buffer pointed to by message. Tag is a programmer-specified identifier that becomes associated with the message, and can be used, for example, to organize the communication streams (e.g. to distinguish two distinct streams of interleaved data). Our examples do not require this, so we will pass in the value 0 for the tag. Comm is the communicator described above, and we will continue to use MPI_COMM_WORLD.

CBoost (C++)FortranFortran 2008Python (mpi4py)

 int MPI_Send
 (
     void *message,           /* reference to data to be sent */
     int count,               /* number of items in message */
     MPI_Datatype datatype,   /* type of item in message */
     int dest,                /* rank of process to receive message */
     int tag,                 /* programmer specified identifier */
     MPI_Comm comm            /* communicator */
 );

template<typename T> void mpi::communicator::send(
  int dest,                  /* rank of process to receive message */ 
  int tag,                          /* programmer specified identified */
  const T & value              /* message */
) const;

 MPI_SEND(MESSAGE, COUNT, DATATYPE, DEST, TAG, COMM, IERR)
 <type> MESSAGE(*)
 INTEGER :: COUNT, DATATYPE, DEST, TAG, COMM, IERR

 MPI_Send(message, count, datatype, dest, tag, comm, ierr)
 TYPE(*), DIMENSION(..), INTENT(IN) :: message
 INTEGER, INTENT(IN) :: count, dest, tag
 TYPE(MPI_Datatype), INTENT(IN) :: datatype
 TYPE(MPI_Comm), INTENT(IN) :: comm
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

# For general Python objects (pickled)
MPI.Intracomm.send(self, obj, int dest, int tag=0)

# For numpy arrays (fast)
MPI.Intracomm.Send(self, buf, int dest, int tag=0)

Note that the datatype argument, specifying the type of data contained in the message buffer, is a variable. This is intended to provide a layer of compatibility between processes that could be running on architectures for which the native format for these types differs. It is possible to register new data types, but for this tutorial we will only use the predefined types provided by MPI. There is a predefined MPI type for each atomic data type in the source language (for C: MPI_CHAR, MPI_FLOAT, MPI_SHORT, MPI_INT, etc. and for Fortran: MPI_CHARACTER, MPI_INTEGER, MPI_REAL, etc.). You can find a full list of these types in the references provided below.

MPI_Recv works in much the same way as MPI_Send. Referring to the function prototypes below, message is now a pointer to an allocated buffer of sufficient size to store count instances of datatype, to be received from process rank. MPI_Recv takes one additional argument, status, which should, in C, be a reference to an allocated MPI_Status structure, and, in Fortran, be an array of MPI_STATUS_SIZE integers or, for mpi_f08, a derived TYPE(MPI_Status) variable. Upon return it will contain some information about the received message. Although we will not make use of it in this tutorial, the argument must be present.

CBoost (C++)FortranFortran 2008Python (mpi4py)

 int MPI_Recv
 (
     void *message,           /* reference to buffer for received data */
     int count,               /* number of items to be received */
     MPI_Datatype datatype,   /* type of item to be received */
     int source,              /* rank of process from which to receive */
     int tag,                 /* programmer specified identifier */
     MPI_Comm comm            /* communicator */
     MPI_Status *status       /* stores info. about received message */
 );

template<typename T> status mpi::communicator::recv(
  int source,                  /* rank of process from which to receive */ 
  int tag,                          /* programmer specified identified */
  const T & value               /* message */
) const;

 MPI_RECV(MESSAGE, COUNT, DATATYPE, SOURCE, TAG, COMM, STATUS, IERR)
 <type> :: MESSAGE(*)
 INTEGER :: COUNT, DATATYPE, SOURCE, TAG, COMM, STATUS(MPI_STATUS_SIZE), IERR

 MPI_Recv(message, count, datatype, source, tag, comm, status, ierr)
 TYPE(*), DIMENSION(..) :: message
 INTEGER, INTENT(IN) :: count, source, tag
 TYPE(MPI_Datatype), INTENT(IN) :: datatype
 TYPE(MPI_Comm), INTENT(IN) :: comm
 TYPE(MPI_Status) :: status
 INTEGER, OPTIONAL, INTENT(OUT) :: ierr

# For general Python objects (pickled)
MPI.Intracomm.recv(self, buf=None, int source=ANY_SOURCE, int tag=ANY_TAG, Status status=None)

# For numpy arrays (fast)
MPI.Intracomm.Recv(self, buf, int source=ANY_SOURCE, int tag=ANY_TAG, Status status=None)

With this simple use of MPI_Send and MPI_Recv, the sending process must know the rank of the receiving process, and the receiving process must know the rank of the sending process. In our example the following arithmetic is useful:

• (rank + 1) % size is the process to send to, and
• (rank + size - 1) % size is the process to receive from.

We can now make the required modifications to our parallel "Hello, world!" program as shown below.

CBoost (C++)FortranFortran 2008Python (mpi4py)

File : phello2.c

#include <stdio.h>
#include <mpi.h>
 
#define BUFMAX 81
 
int main(int argc, char *argv[])
{
     char outbuf[BUFMAX], inbuf[BUFMAX];
     int rank, size;
     int sendto, recvfrom;
     MPI_Status status;
 
     MPI_Init(&argc, &argv);
     MPI_Comm_rank(MPI_COMM_WORLD, &rank);
     MPI_Comm_size(MPI_COMM_WORLD, &size);
 
     sprintf(outbuf, "Hello, world! from process %d of %d", rank, size);
 
     sendto = (rank + 1) % size;
     recvfrom = (rank + size - 1) % size;
 
     MPI_Send(outbuf, BUFMAX, MPI_CHAR, sendto, 0, MPI_COMM_WORLD);
     MPI_Recv(inbuf, BUFMAX, MPI_CHAR, recvfrom, 0, MPI_COMM_WORLD, &status);
 	
     printf("[P_%d] process %d said: \"%s\"]\n", rank, recvfrom, inbuf);
 
     MPI_Finalize();
     return(0);
}

File : phello.cpp

#include <iostream>
#include <string>
#include <boost/mpi.hpp>

using namespace std;
using namespace boost;

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

    string outmessage = "Hello, world! from process " + to_string(rank) + " of " + to_string(size);
    string inmessage;
    int sendto = (rank + 1) % size;
    int recvfrom = (rank + size - 1) % size;

    cout << outmessage << endl;

    world.send(sendto,0,outmessage);
    world.recv(recvfrom,0,inmessage);

    cout << "[P_" << rank << "] process " << recvfrom << " said: \"" << inmessage << "\"" << endl;
    return 0;
}

File : phello2.f90

program phello2

     implicit none
     use mpi
     integer, parameter :: BUFMAX=81
     character(len=BUFMAX) :: outbuf, inbuf, tmp
     integer :: rank, num_procs, ierr
     integer :: sendto, recvfrom
     integer :: status(MPI_STATUS_SIZE)
 
     call MPI_INIT(ierr)
     call MPI_COMM_RANK(MPI_COMM_WORLD, rank, ierr)
     call MPI_COMM_SIZE(MPI_COMM_WORLD, num_procs, ierr)
 
     outbuf = 'Hello, world! from process ' 
     write(tmp,'(i2)') rank
     outbuf = outbuf(1:len_trim(outbuf)) // tmp(1:len_trim(tmp))
     write(tmp,'(i2)') num_procs
     outbuf = outbuf(1:len_trim(outbuf)) // ' of ' // tmp(1:len_trim(tmp))
 
     sendto = mod((rank + 1), num_procs)
     recvfrom = mod((rank + num_procs - 1), num_procs)
 
     call MPI_SEND(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD, ierr)
     call MPI_RECV(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status, ierr)
 
     print *, 'Process', rank, ': Process', recvfrom, ' said:', inbuf
 
     call MPI_FINALIZE(ierr)
 
end program phello2

File : phello2.f90

program phello2

     implicit none
     use mpi_f08
     integer, parameter :: BUFMAX=81
     character(len=BUFMAX) :: outbuf, inbuf, tmp
     integer :: rank, num_procs
     integer :: sendto, recvfrom
     type(MPI_Status) :: status
 
     call MPI_Init()
     call MPI_Comm_rank(MPI_COMM_WORLD, rank)
     call MPI_Comm_size(MPI_COMM_WORLD, num_procs)
 
     outbuf = 'Hello, world! from process ' 
     write(tmp,'(i2)') rank
     outbuf = outbuf(1:len_trim(outbuf)) // tmp(1:len_trim(tmp))
     write(tmp,'(i2)') num_procs
     outbuf = outbuf(1:len_trim(outbuf)) // ' of ' // tmp(1:len_trim(tmp))
 
     sendto = mod((rank + 1), num_procs)
     recvfrom = mod((rank + num_procs - 1), num_procs)
 
     call MPI_Send(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD)
     call MPI_Recv(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status)
 
     print *, 'Process', rank, ': Process', recvfrom, ' said:', inbuf
 
     call MPI_Finalize()
 
end program phello2

File : phello2.py

from mpi4py import MPI

comm = MPI.COMM_WORLD
rank = comm.Get_rank()
size = comm.Get_size()

outbuf = "Hello, world! from process %d of %d" % (rank, size)
 
sendto = (rank + 1) % size;
recvfrom = (rank + size - 1) % size;
 
comm.send(outbuf, dest=sendto, tag=0)
inbuf = comm.recv(source=recvfrom, tag=0)
 	
print('[P_%d] process %d said: "%s"]' % (rank, recvfrom, inbuf))

Compile this program and run it using 2, 4, and 8 processes. While it certainly seems to be working as intended, there is a hidden problem here. The MPI standard does not guarantee that MPI_Send returns before the message has been delivered. Most implementations buffer the data from MPI_Send and return without waiting for it to be delivered. But if it were not buffered, the code we've written would deadlock: Each process would call MPI_Send and then wait for its neighbour process to call MPI_Recv. Since the neighbour would also be waiting at the MPI_Send stage, they would all wait forever. Clearly there is buffering in the libraries on our systems since the code did not deadlock, but it is poor design to rely on this. The code could fail if used on a system in which there is no buffering provided by the library. Even where buffering is provided, the call might still block if the buffer fills up.

[~]$ mpicc -Wall phello2.c -o phello2
[~]$ mpirun -np 4 ./phello2
[P_0] process 3 said: "Hello, world! from process 3 of 4"]
[P_1] process 0 said: "Hello, world! from process 0 of 4"]
[P_2] process 1 said: "Hello, world! from process 1 of 4"]
[P_3] process 2 said: "Hello, world! from process 2 of 4"]

Safe MPI

The MPI standard defines MPI_Send and MPI_Recv to be blocking calls. This means MPI_Send will not return until it is safe for the calling module to modify the contents of the provided message buffer. Similarly, MPI_Recv will not return until the entire contents of the message are available in the message buffer the caller provides.

It should be obvious that whether or not the MPI library provides buffering does not affect receive operations. As soon as the data is received, it will be placed directly in the message buffer provided by the caller and MPI_Recv will return; until then the call will be blocked. MPI_Send on the other hand need not block if the library provides an internal buffer. Once the message is copied out of the original data location, it is safe for the user to modify that location, so the call can return. This is why our parallel "Hello, world!" program doesn't deadlock as we have implemented it, even though all processes call MPI_Send first. Since the buffering is not required by the MPI standard and the correctness of our program relies on it, we refer to such a program as unsafe MPI.

A safe MPI program is one that does not rely on a buffered implementation in order to function correctly. The following pseudo-code fragments illustrate this concept:

Deadlock

 ...
    if (rank == 0)
    {
        MPI_Recv(from 1);
        MPI_Send(to 1);
    }
    else if (rank == 1)
    {
        MPI_Recv(from 0);
        MPI_Send(to 0);
    }
...

Receives are executed on both processes before the matching send; regardless of buffering, the processes in this MPI application will block on the receive calls and deadlock.

Unsafe

...
    if (rank == 0)
    {
        MPI_Send(to 1);
        MPI_Recv(from 1);
    }
    else if (rank == 1)
    {
        MPI_Send(to 0);
        MPI_Recv(from 0);
    }
...

This is essentially what our parallel "Hello, world!" program was doing, and it may work if buffering is provided by the library. If the library is unbuffered, or if messages are simply large enough to fill the buffer, this code will block on the sends, and deadlock. To fix that we will need to do the following instead:

Safe

...
    if (rank == 0)
    {
        MPI_Send(to 1);
        MPI_Recv(from 1);
    }
    else if (rank == 1)
    {
        MPI_Recv(from 0);
        MPI_Send(to 0);
    }
...

Even in the absence of buffering, the send here is paired with a corresponding receive between processes. While a process may block for a time until the corresponding call is made, it cannot deadlock.

How do we rewrite our "Hello, World!" program to make it safe? A common solution to this kind of problem is to adopt an odd-even pairing and perform the communication in two steps. Since in our example communication is a rotation of data one rank to the right, we should end up with a safe program if all even ranked processes execute a send followed by a receive, while all odd ranked processes execute a receive followed by a send. The reader can easily verify that the sends and receives are properly paired avoiding any possibility of deadlock.

CBoost (C++)FortranFortran 2008Python (mpi4py)

File : phello3.c

#include <stdio.h>
#include <mpi.h>

#define BUFMAX 81

int main(int argc, char *argv[])
{
    char outbuf[BUFMAX], inbuf[BUFMAX];
    int rank, size;
    int sendto, recvfrom;
    MPI_Status status;


    MPI_Init(&argc, &argv);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &size);

    sprintf(outbuf, "Hello, world! from process %d of %d", rank, size);

    sendto = (rank + 1) % size;
    recvfrom = ((rank + size) - 1) % size;

    if (!(rank % 2))
    {
        MPI_Send(outbuf, BUFMAX, MPI_CHAR, sendto, 0, MPI_COMM_WORLD);
        MPI_Recv(inbuf, BUFMAX, MPI_CHAR, recvfrom, 0, MPI_COMM_WORLD, &status);
    }
    else
    {
        MPI_Recv(inbuf, BUFMAX, MPI_CHAR, recvfrom, 0, MPI_COMM_WORLD, &status);
        MPI_Send(outbuf, BUFMAX, MPI_CHAR, sendto, 0, MPI_COMM_WORLD);
    }

    printf("[P_%d] process %d said: \"%s\"]\n", rank, recvfrom, inbuf);

    MPI_Finalize();

    return(0);
}

File : phello3.cpp

#include <iostream>
#include <string>
#include <boost/mpi.hpp>

using namespace std;
using namespace boost;

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

    string outmessage = "Hello, world! from process " + to_string(rank) + " of " + to_string(size);
    string inmessage;
    int sendto = (rank + 1) % size;
    int recvfrom = (rank + size - 1) % size;

    cout << outmessage << endl;

    if (!(rank % 2)) {
        world.send(sendto,0,outmessage);
        world.recv(recvfrom,0,inmessage);
    }
    else {
        world.recv(recvfrom,0,inmessage);
        world.send(sendto,0,outmessage);
    }

    cout << "[P_" << rank << "] process " << recvfrom << " said: \"" << inmessage << "\"" << endl;
    return 0;
}

File : phello3.f90

program phello3


    implicit none
    use mpi

    integer, parameter :: BUFMAX=81
    character(len=BUFMAX) :: outbuf, inbuf, tmp
    integer :: rank, num_procs, ierr
    integer :: sendto, recvfrom
    integer :: status(MPI_STATUS_SIZE)

    call MPI_INIT(ierr)
    call MPI_COMM_RANK(MPI_COMM_WORLD, rank, ierr)
    call MPI_COMM_SIZE(MPI_COMM_WORLD, num_procs, ierr)

    outbuf = 'Hello, world! from process '
    write(tmp,'(i2)') rank
    outbuf = outbuf(1:len_trim(outbuf)) // tmp(1:len_trim(tmp))
    write(tmp,'(i2)') num_procs
    outbuf = outbuf(1:len_trim(outbuf)) // ' of ' // tmp(1:len_trim(tmp))

    sendto = mod((rank + 1), num_procs)
    recvfrom = mod(((rank + num_procs) - 1), num_procs)

    if (MOD(rank,2) == 0) then
        call MPI_SEND(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD, ierr)
        call MPI_RECV(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status, ierr)
    else
        call MPI_RECV(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status, ierr)
        call MPI_SEND(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD, ierr)
    endif

    print *, 'Process', rank, ': Process', recvfrom, ' said:', inbuf

    call MPI_FINALIZE(ierr)

end program phello3

File : phello3.f90

program phello3


    implicit none
    use mpi_f08

    integer, parameter :: BUFMAX=81
    character(len=BUFMAX) :: outbuf, inbuf, tmp
    integer :: rank, num_procs
    integer :: sendto, recvfrom
    type(MPI_Status) :: status

    call MPI_Init()
    call MPI_Comm_rank(MPI_COMM_WORLD, rank)
    call MPI_Comm_size(MPI_COMM_WORLD, num_procs)

    outbuf = 'Hello, world! from process '
    write(tmp,'(i2)') rank
    outbuf = outbuf(1:len_trim(outbuf)) // tmp(1:len_trim(tmp))
    write(tmp,'(i2)') num_procs
    outbuf = outbuf(1:len_trim(outbuf)) // ' of ' // tmp(1:len_trim(tmp))

    sendto = mod((rank + 1), num_procs)
    recvfrom = mod(((rank + num_procs) - 1), num_procs)

    if (MOD(rank,2) == 0) then
        call MPI_Send(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD)
        call MPI_Recv(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status)
    else
        call MPI_RECV(inbuf, BUFMAX, MPI_CHARACTER, recvfrom, 0, MPI_COMM_WORLD, status)
        call MPI_SEND(outbuf, BUFMAX, MPI_CHARACTER, sendto, 0, MPI_COMM_WORLD)
    endif

    print *, 'Process', rank, ': Process', recvfrom, ' said:', inbuf

    call MPI_Finalize()

end program phello3

File : phello3.py

from mpi4py import MPI

comm = MPI.COMM_WORLD
rank = comm.Get_rank()
size = comm.Get_size()

outbuf = "Hello, world! from process %d of %d" % (rank, size)

sendto = (rank + 1) % size;
recvfrom = ((rank + size) - 1) % size;

if rank % 2 == 0:
    comm.send(outbuf, dest=sendto, tag=0)
    inbuf = comm.recv(source=recvfrom, tag=0)
else:
    inbuf = comm.recv(source=recvfrom, tag=0)
    comm.send(outbuf, dest=sendto, tag=0)

print('[P_%d] process %d said: "%s"]' % (rank, recvfrom, inbuf))

Is there still a problem here if the number of processors is odd? It might seem so at first as process 0 (which is even) will be sending while process N-1 (also even) is trying to send to 0. But process 0 is originating a send that is correctly paired with a receive at process 1. Since process 1 (odd) begins with a receive, that transaction is guaranteed to complete. When it does, process 0 will proceed to receive the message from process N-1. There may be a (very small!) delay, but there is no chance of a deadlock. [~]$ mpicc -Wall phello3.c -o phello3

[~]$ mpirun -np 16 ./phello3
[P_1] process 0 said: "Hello, world! from process 0 of 16"]
[P_2] process 1 said: "Hello, world! from process 1 of 16"]
[P_5] process 4 said: "Hello, world! from process 4 of 16"]
[P_3] process 2 said: "Hello, world! from process 2 of 16"]
[P_9] process 8 said: "Hello, world! from process 8 of 16"]
[P_0] process 15 said: "Hello, world! from process 15 of 16"]
[P_12] process 11 said: "Hello, world! from process 11 of 16"]
[P_6] process 5 said: "Hello, world! from process 5 of 16"]
[P_13] process 12 said: "Hello, world! from process 12 of 16"]
[P_8] process 7 said: "Hello, world! from process 7 of 16"]
[P_7] process 6 said: "Hello, world! from process 6 of 16"]
[P_14] process 13 said: "Hello, world! from process 13 of 16"]
[P_10] process 9 said: "Hello, world! from process 9 of 16"]
[P_4] process 3 said: "Hello, world! from process 3 of 16"]
[P_15] process 14 said: "Hello, world! from process 14 of 16"]
[P_11] process 10 said: "Hello, world! from process 10 of 16"]

Note that many frequently-occurring communication patterns have been captured in the collective communication functions of MPI. If there is a collective function that matches the communication pattern you need, you should use it instead of implementing it yourself with MPI_Send and MPI_Recv.

MPI compiler-wrappers

The packages of MPI libraries typically provide so-called "wrappers" for compilers. These are not compilers themselves but call other compilers and at the same time make sure to pass MPI-specific flags to the compiler.

Typically they are called (this list does not attempt to be complete):

• mpicc for C
• mpicxx for C++
• mpifort and/or mpif77 and/or mpif90 for Fortran

With Open MPI, these wrappers have the option --showme that will print which compiler will be called, and which compiler-options will be added.

Note that our software stack contains several modules of the same Open MPI package (e.g. openmpi/4.0.3) that have been built with different compilers, and with and without CUDA support. The MPI compiler-wrappers will always use the compiler and compiler version that you have loaded with the module load command.

For example, if you have loaded the intel/2020.1.217 and openmpi/4.0.3 modules:

[name@server ~]$ mpicc -showme
icc -I/cvmfs/…/intel2020/openmpi/4.0.3/include -L/cvmfs/…/intel2020/openmpi/4.0.3/lib -lmpi

If you have loaded gcc/9.3.0 and openmpi/4.0.3 modules:

[name@server ~]$ mpicc -showme
gcc -I/…/gcc9/openmpi/4.0.3/include -L/cvmfs/…/gcc9/openmpi/4.0.3/lib -lmpi

Comments and Further Reading

This tutorial presented some of the key design and syntax concepts associated with MPI programming. There is still a wealth of material to be considered in designing any serious parallel program, including but not limited to:

• MPI_Send/MPI_Recv variants (buffered, non-blocking, synchronous, etc.)
• collective communications (reduction, broadcast, barrier, scatter, gather, etc.)
• derived data types
• communicators and topologies
• one-sided communication and other features of MPI-2
• efficiency issues
• parallel debugging
• Tutorial on Boost MPI (in French)

Selected references

• William Gropp, Ewing Lusk, and Anthony Skjellum. Using MPI: Portable Parallel Programming with the Message-Passing Interface (2e). MIT Press, 1999.
  • Comprehensive reference covering Fortran, C and C++ bindings
• Peter S. Pacheco. Parallel Programming with MPI. Morgan Kaufmann, 1997.
  • Easy to follow tutorial-style approach in C.
• Blaise Barney. Message Passing Interface (MPI). Lawrence Livermore National Labs.
• Wes Kendall, Dwaraka Nath, Wesley Bland et al. mpitutorial.com.
• Various authors; IDRIS. Formation "MPI" (en français).

Message Passing Interface