Summer Workshops 2006:

Introduction to High Performance Computing 

Instructor: John Pormann, PhD
Dates: June 6th, 7th, and 8th
Location: Duke University, 306 North Building
Cost: $50  (includes lunch and course materials)

Description:
The focus of this workshop will be on techniques and tools for  assessing and improving the performance of computer programs. We will begin with some of the relevant details of modern computer  architecture, especially cluster and parallel systems. Speed and memory optimization will be covered, as well as file system and  network performance.  We will walk through some simplified examples  to show how different tools can be used to determine where a problem  exists and also give options for how to improve the program. A brief  introduction to parallelism will be given, with a focus on how to  extract the parallelism within real applications.

The workshop will include several lab sessions and participants are  invited to bring their own research code for use in the labs.

Prerequisites:
Students should have a working knowledge of either C/C++ or FORTRAN.

Other Information:
Accounts will be provided on a small cluster computing system for the duration of the workshop.  Students will be asked to provide their Duke University NetID as part of registration.  Participants without NetIDs will be provided with local accounts.

Registration website: http://events.duke.edu/csemihpc

Programming with Message Passing Interface (MPI)

Instructor: Bill Rankin, PhD
Dates: June 21st, 22nd, 23rd
Location: Duke University, 306 North Building
Cost: $50 (includes lunch and course materials)

Description:
The focus of this workshop will be the application of parallel programming and the practical use of high performance computational clusters to solve research problems in various fields.  MPI, the Message Passing Interface, is a widely used and highly portable standard library that provides programmers with a method of implementing parallel algorithms.

The emphasis of this workshop is on the use of MPI library calls to solve real programming problems. The workshop will begin with a general overview of the practice of parallel programming and some general approaches to algorithm development. We will then cover the core routines of the MPI library and show examples of their use. The goal of the course is to provide the students with the basic knowledge necessary to begin implementing parallel distributed programs.

Prerequisites:
Students should have a working knowledge of either C/C++ or FORTRAN.

Other Information:
Accounts will be provided on a small cluster computing system for the duration of the workshop.  Students will be asked to provide their Duke University NetID as part of registration.  Participants without NetIDs will be provided with local accounts.

Registration website: http://events.duke.edu/csemmpi