|> home > support > training > HPCxMPIIO|
Parallel IO using MPI-IO
One of the greatest challenges to running parallel applications on large numbers of processors is how to handle file IO. The standard Unix IO routines are not designed with parallelism in mind, and multiple processes simultaneously writing to the same file can give unpredictable results.
The IO part of the MPI-2 standard, which is now widely implemented, gives programmers access to efficient parallel IO in a portable fashion. However, there are a large number of different MPI-IO routines available and they can be difficult to use in practice. Specifically, they require a more detailed understanding of MPI derived datatypes than is required for most MPI programs.
Despite its apparent complexity, MPI-IO follows a very straightforward model. It is well worth the effort of learning how to use MPI-IO as it enables parallel applications to use a single file format regardless of the number of processes being used. In other words, the format of the parallel file is identical to that produced by a serial program.
Venue: Training Room B43a, B Block, STFC Daresbury Laboratory, UK
This one-day course will cover:
The course will contain a number of practical sessions working on programming exercises. These are designed to illustrate the basic methods and to act as useful templates for using MPI-IO in real applications codes.
The course assumes a good understanding of basic MPI programming in Fortran, C or C++. Knowledge of derived datatypes would be useful but not essential.
To apply for a place on this course please complete this form.
For any additional information please the HPCx helpdesk.
|http://www.hpcx.ac.uk/support/training/HPCxMPIIO.html||contact email - firstname.lastname@example.org||© UoE HPCX Ltd|