Abstract
The most common type of high-performance parallel computer is a distributed memory computer: a computer that consists of many processors, each with their own individual memory, that can only access the data stored by other processors by passing messages across a network. This chapter serves as an introduction to the Message Passing Interface (MPI), which is a widely used library of code for writing parallel programs on distributed memory architectures. Although the MPI libraries contain many different functions, basic code may be written using only a very small subset of these functions. By providing a basic guide to these commonly used functions you will be able to write simple MPI programs, edit MPI programs written by other programmers, and understand the function calls when using a scientific library built on MPI.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
The Portable Extensible Toolkit for Scientific Computing (PETSc, pronounced “pet see”) is a library providing functionality for the solution of linear and nonlinear systems of equations on both sequential and parallel architectures.
- 2.
There are several programming libraries which allow the programmer access to a distributed shared memory computer where machines over a network act as if they were part on one contiguous system. There has, however, not been wide-spread use of these libraries at the time of writing.
- 3.
MPI implementations vary in how they return console input from the individual processes to the console from which the program was launched. Even when flush is called on the cout stream it may still be the case that the MPI machinery is buffering output.
- 4.
MPI::ANY_TAG and MPI::ANY_SOURCE are C++ names for these wild-card values. Many codes use the interchangeable C names: MPI_ANY_TAG and MPI_ANY_SOURCE.
- 5.
Note that these are the C++ object names for these types—they are also called synonymously by their C names: MPI_BOOL, MPI_CHAR, MPI_INT and MPI_DOUBLE.
- 6.
There are a few standard ways of getting data to file from a parallel program: concentration, where one process does all the writing, as suggested above; round-robin where processes take it in turns to open and close the same file; parallel file libraries such as MPI’s MPIIO; and separate files where each process writes data to different places to be re-assembled later. The choice of output method is largely dependent on the data structure and size.
References
The Message-Passing Interface (MPI)
Gropp, W., Lusk, E., Skjellum, A.: Using MPI: Portable Parallel Programming with the Message-Passing Interface, 2nd edn. Massachusetts Institute of Technology Press, Cambridge (1999)
Gropp, W., Lusk, E., Thakur, R.: Using MPI-2: Advanced Features of the Message-Passing Interface. Massachusetts Institute of Technology Press, Cambridge (1999)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2012 Springer-Verlag London Limited
About this chapter
Cite this chapter
Pitt-Francis, J., Whiteley, J. (2012). An Introduction to Parallel Programming Using MPI. In: Guide to Scientific Computing in C++. Undergraduate Topics in Computer Science. Springer, London. https://doi.org/10.1007/978-1-4471-2736-9_11
Download citation
DOI: https://doi.org/10.1007/978-1-4471-2736-9_11
Publisher Name: Springer, London
Print ISBN: 978-1-4471-2735-2
Online ISBN: 978-1-4471-2736-9
eBook Packages: Computer ScienceComputer Science (R0)