Parallel Computing in Undergraduate Education

Joint Institute for Computational Science, UTK/ORNL

Course Description

Theme of the Course

With the emergence of parallel supercomputers, the art of computational science has leaped to a new era. Problems that were not tractable before are now being solved. The huge amount of computational resources have been used for searching new drugs, building efficient aeroplanes, discovering new composite materials, predicting snow storms, and understanding human genetic structures. Although the physics remains unchanged, revision of existing algorithms for numerical simulations on parallel computers is most likely inevitable. The dynamic computer industry has moved so swiftly that working on parallel computers is no long limited to researchers at prestigious institutions or national laboratories. The envolope of high performance computing has gradually expanded to engulf ordinary engineers and industrial product developers. Many references aimed at educating students in high-performace computing can be found on the world wide web. However, an intuitive web-based undergraduate course that would help faculty members to deliver the science of high performance computing to their students has yet to be composed. The purpose of the course is to help faculty members of natural sciences and engineering to integrate principle ideas of high-performance computing into their undergraduate curricula. The materials offered in this workshop serve primarily as a guideline to help faculty members design and organize information of high-performance computing in a structural and intuitive manner. Since undergraduate students are the primary audience of the subject, advanced details of the presented topics will be neglected. Emphasis will be spent on implementation of practical examples using numerical algoithms which are well known to most junior and senior college students. Faculty members using the course materials should incorporate the appropriate portions of the workshop according to their needs and levels of appreciations of their students. Afterall, whatever amount of knowlegde students derive from the subject is the final measure of the achievement of the educators.

Hardware and Software

A network of personal computers will be the primary computing platform to support the course. At least two 200MHz Pentium Pro PCs connected with Ethernet or other devices will be needed and utilized. Versions of PVM and MPI inplementations are freely available for PCs running Linux. Site licences of HPF on PCs running Linux may be purchased from Portland Group. Thus, Linux will have to be installed on the network of PCs. Details of hardware requirements and software installations will be provided as supplements to the course for the attendees of the workshop. Access to a network of Sun workstations at UTK and educational accounts at CTC, PSC, or NCSA will be solicited as necessary.

Course layout

The course is composed of several modules. Each module covers a specific topic in detail. The modules are classified under three sections:

Overview and Basics

Modules give an overview of computing in general. Some fundamental principles of scientific computing are introduced. The basic architetures of various parallel computers are examined. Some general theories and principles of parallel computing are also discussed.

Techniques of Parallel Programming

Data parallelism and message passing programming techniques are introduced. Two examples, integration and matrix multiplication, are used to introduce and illustrate the implementation of parallel programs. Parallel programs written with HPF, PVM, and MPI are shown and discussed. Hands-on exercises will be provided for participants to work on a network of workstations.

Computing Applications

Examples of computing applications are examineds. Topics of applications consist primarily of numerical algorithms that are familiar to senior level college students. Topics include Gaussian elimination, iterative techniques for solving systems of equations, methods for solving ordinary and partial differential equations, sorting algorithms, searching algorithms, etc..

References :

BOOKS :
- Introduction to Parallel Computing: Design and Analysis of Algorithm by Vipin Kumar, Ananth Grama, Anshul Gupta, and George Karypis
- An Introduction to Parallel Algorithm by Joseph JaJa
- A Scientist's and Engineer's Guide to Wprkstations and Supercomputers by Rubin Landau and Paul Fink, Jr.
- Solving Linear Systems on Vector and Shared Memory Computers by Jack Dongarra, Iain Duff, Danny Sorensen, and Henk van der Vorst.
- Parallel and Distributed Computation: Numerical Methods by Dimitri Bertsekas and John N. Tsitsiklis.
- Parallel Computing: Theory and Practice by Michael Quinn
- Using MPI by William Gropp
- Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods by R. Barrett, M. Berry, T. Chan, J. Demmel, J. Donato, J. Dongarra, V. Eijkhout, R. Pozo, C. Rammie, and H. van der Vorst.
E-BOOKS
LECTURES :
WORKSHOPS:
MISCELLANEOUS: