Message Passing Interface Lectures 2002

Description

This series of lectures is an introduction to the Message Passing Interface (MPI), a standard library of subroutines (Fortran) or function calls (C) that can be used to implement a message passing program. MPI allows the coordination of a program running as multiple processes in a distributed memory environment, yet is flexible enough to be used in a shared memory system. MPI programs can be used and compiled on a wide variety of parallel computers or a cluster of workstations over a network.

The standardization of the MPI library is one of its most powerful features. This means that the parallel programmer can write code containing MPI subroutine and function calls that will work on *any* machine on which the MPI library is installed without having to make changes in the this code. At present MPI can be found on all parallel supercomputers on which research-level computation is being performed.

• Parallel Computer Architectures and Models of Parallel Programming
• Getting Started with MPI
• MPI Program Structure
• Point-to-Point Communication
• Communication Modes
• Derived Datatypes
• Collective Communications
• Communicators
• Virtual Topologies
• Program Performance

The lectures take place on Wednesdays 3:00-4:00 pm in LSK 301 (IAM seminar room), starting on January 16th and ending on March 27th, 2002. All lectures are given by Roman Baranowski (roman@chem.ubc.ca), except for Lecture 6, presented by Henryk Modzelewski from the Geophysical Disaster and Computational Fluid Dynamics Centre at UBC. No prior experience with MPI or parallel programming is required. However, an understanding of computer programming is necessary.

Transcripts

 Lecture 1: Parallel Architectures   [ pdf ]

 Lecture 2: Point to Point Communication   [ pdf ]

 Lectures 3-4: Global Communication and Derived Data Types   [ pdf ]

 Lecture 5: Communicators and Virtual Topologies   [ pdf ]

 Lecture 6: GDCFD Parallel Computing: Hardware and Software   [ pdf ]

 Program examples: Fortran 90 + MPI

Additional Resources

• IBM Redbooks RS/6000 SP: Practical MPI Programming
• EPCC Training and Education Documents
• MPI Tutorials on the web
• Geophysical Disaster and Computational Fluid Dynamics Centre at UBC provides access to IBM Linux Super-Cluster (Monster)