Supercomputing

• HPC Support
• PI Portal
• Get a New OSC Account
• Request Additional
   Resources
• Request a Classroom
   Project
• FAQ
• Computing
  Environments
• Available Hardware
• Training
• Accounts
• Available Software
• Statewide Software
• Manuals
• Consult
• Notices
• Policies

Related Links

• Statewide Users Group
• Blue Collar Computing
• Ralph Regula School
  of Computational
  Science
• BALE
• Interface Lab

Multilevel Parallel Programming

Description:

This course is intended as an introduction to multilevel parallel programming, a style of parallel programming in which both message passing techniques (such as MPI) and shared memory techniques (such as OpenMP or pthreads) are used. This allows high performance codes to best take advantage of the distributed/shared memory architectures of modern parallel supercomputers such as SGI Origins, IBM SP3s, and clusters of commodity SMP systems. Users will learn how to apply multilevel parallel programming techniques to their problems of interest, how to optimize these techniques for different architectures, and how to avoid potential problems.

Topics covered will include:

• Introduction to distributed/shared memory architectures
• Overview of message passing techniques
• Overview of shared memory techniques
• Mixing message passing and shared memory
• Optimization techniques
• Examples

Target Audience:

Users of distributed/shared memory systems such as the SGI Origin 2000 or the OSC/SGI Beowulf Cluster.
 

Prerequisites:

Familiarity with a programming language (preferably Fortran or C) and parallel programming using either MPI or OpenMP is required. Knowledge of both MPI and OpenMP is helpful.
 

Method of Delivery:

Lecture and hands-on laboratory.

Handouts:

May 1 - May 2, 2001, PDF, Jim Guiliani (jimg@osc.edu)