The Cray Message Passing Toolkit and Message Passing Toolkit for IRIX (MPT) is a software package that supports interprocess data exchange for applications that use concurrent, cooperating processes on a single host or on multiple hosts. Data exchange is done through message passing, which is the use of library calls to request data delivery from one process to another or between groups of processes.
The MPT 1.3 package contains the following components and the appropriate accompanying documentation:
Parallel Virtual Machine (PVM)
Message Passing Interface (MPI)
Logically shared, distributed memory (SHMEM) data-passing routines
The Message Passing Interface (MPI) is a standard specification for a message passing interface, allowing portable message passing programs in Fortran and C languages.
This chapter provides an overview of the MPI software that is included in the toolkit, a description of the basic MPI components, and a list of general steps for developing an MPI program. Subsequent chapters address the following topics:
Building MPI applications
Using mpirun to execute applications
Setting environment variables
Launching programs with NQE
MPI is a standard specification for a message passing interface, allowing portable message passing programs in Fortran and C languages. MPI was created by the Message Passing Interface Forum (MPIF). MPIF is not sanctioned or supported by any official standards organization. Its goal was to develop a widely used standard for writing message passing programs. Silicon Graphics supports implementations of MPI that are released as part of the Message Passing Toolkit. These implementations are available on IRIX, UNICOS, and UNICOS/mk systems. The MPI standard is available from the IRIS InSight library (for customers who have access to that tool), and is documented online at the following address:
http://www.mcs.anl.gov/mpi |
On UNICOS and UNICOS/mk systems, the MPI library is provided as a statically linked set of objects (a file with a name that ends in .a). On IRIX systems, the MPI library is provided as a dynamic shared object (DSO) (a file with a name that ends in .so). The basic components that are necessary for using MPI are the libmpi.a library (or libmpi.so for IRIX systems), the include files, and the mpirun(1) command.
For UNICOS/mk systems, in addition to the libmpi.a library, the libpmpi.a library is provided for profiling support. For UNICOS and IRIX systems, profiling support is included in the libmpi.a and libmpi.so libraries, respectively. Profiling support replaces all MPI_ Xxx prototypes and function names with PMPI_Xxx entry points.
To develop a program that uses MPI, you must perform the following steps:
Procedure 1-1. Steps for MPI program development
Add MPI function calls to your application for MPI initiation, communications, and synchronization. For descriptions of these functions, see the online man pages or Using MPI: Portable Parallel Programming with the Message-Passing Interface or the MPI standard specification.
Build programs for the systems that you will use, as described in Chapter 2, “Building MPI Applications”.
Execute your program by using one of the following methods:
mpirun(1) (available on all systems (see Chapter 3, “Using mpirun to Execute Applications”))
Execute directly (available on UNICOS/mk systems (see “Executing UNICOS/mk Programs Directly” in Chapter 3))
Note: On IRIX systems, for information on how to execute MPI programs across more than one host or how to execute MPI programs that consist of more than one executable file, see “Building Applications on IRIX Systems” in Chapter 2.