Teaching Parallel Computing to Undergraduates at Capital University

1
Teaching Parallel Computing to Undergraduates at
Capital University

• Ignatios E. VakalisDepartment of Math/CSCapital
  UniversityColumbus OH 43209ivakalis_at_capital.edu

2
Is Parallel Computing Important?

• Large scale scientific problems(Grant
  Challenges)
• Large data-base applications
• Introduces fundamental issues in
  CS(synchronization, threading, non-determinism)
• Cheap (Large number of processors, fast
  interconnection network)

3
Types of (Existing) Curricula

• One course in parallel computing
  (upper division, undergraduate - graduate)
• No course in parallel computing
  (large number of liberal arts undergraduate
  institutions)
• Integrate parallel computing concepts throughout
  the undergraduate curriculum (WOW! Is that
  possible? Stay tuned for the panel discussion.)

4
Parallel Computing at Capital University

• Obtained a grant (NSF - ILI DUE 9551259) to
  establish the Advanced Computational Laboratory
• Advanced Computational Laboratory
• 11 Ultra -10 Sun workstations
• Networked with FastEthernet (100 Mbits/sec)
• 3 SGI -O2 workstations

5
Advanced Computational Lab

• Software for Parallel Computing
• PVM
• MPI
• Parallaxis
• Other software suites
• Maple
• Matlab
• AVS
• NAG

6
CURRICULUM

• Typical Undergraduate CS curriculum
• Two courses in Parallel Computing (CS 376, 476)
• Two courses in Computer Graphics (CS 377, 477)
• Emphasis on undergraduate research
• Junior - senior seminar experience
• Projects in parallel computing and computer
  graphics
• Presentations in state and national meetings
  (NCUR)

7
Introduction to Parallel Computing (CS 376)

• Required course for all CS majors
• Textbook Distributed and Parallel computing by
  H. El-Rewini, T. Lewis, Prentice Hall, 1998
• Prerequisites Data Structures and Algorithms
• Typical class size 20 CS majors,
  5-10 math/science majors

8
Course Outline

• What is Parallel Computing
• next generation
• network computing
• WWW, Client-server
• Performance Measures
• Amdals Law, other measures
• limits of parallelism
• Benchmarking

9

• Distributed and Parallel processors
• Interconnection networks
• (static, dynamic, topologies)
• Programming paradigms (MIMD, SIMD, SM, VP)
• Cache design in SM machines
• The PRAM Model
• Overview
• Analysis of Algorithms
• Matrix multiplication, sorting, searching

10

• Message passing Algorithms
• Overview
• PVM (programming assignments)
• MPI (programming assignments)
• Execute code in ACL and OSC (Cray T94,
• Cray T3E, SGI Origin 2000)
• SIMD data- parallel programming
• Programming model
• Virtual machine
• Programming is parallaxis (simulator)

11
High Performance Computing(CS 476)

• Elective course for CS, Math and Science majors.
• Textbook Introduction to High Performance
  Scientific Computing, by L. Fosdick, et. al.,
  MIT Press, 1996
• Prerequisites Parallel Computing (CS 376)
• Typical Size 10 CS/Math/Science majors

12
Course Outline

• Emphasis on Scientific (Numerical) applications
• Selected topics from Numerical Analysis
• Tools
• Matlab
• AVS (Scientific visualization)
• Issues related to the efficient use of cache
• Improving code performance

13

• Programming with MPI (review from CS376)
• Use of CRAY T94, CRAY T3E, SGI ORIGIN 2000
• Ohio Supercomputer Center
• Introduction to OpenMP
• Multi-threaded applications on a SM environment
• Programming the SGI ORIGIN 2000 (OSC)
• Scaled Vector Processing
• Concepts of vector processing
• Programming the SV1 at OSC

14
Projects

• Solving the wave equation (vibrating string) in
  parallel.
• Model (continuous, disretization)
• Parallel implementation on DM machines
• Parallel Monte Carlo Methods
• Molecular Dynamics Simulations
• Hard spheres model
• Lenorad-Jones intereactions
• Model construction
• Implementation on DM and VP models
• Visualization (AVS, MATLAB)

15
Other Courses

• Computational Science I
• Required by all science and math majors
• Brief overview of Grand Challenge problems
• Reading assignments on the use of parallel
• computing in solving science problems
• Seminar Experience
• Students select to do a research project in
  the area of parallel computing
• Parallel ray tracing and parallel rendering
• Parallel image processing
• Parallel stereo-matching algorithms

16
Present --gt Future

• Parallel computing concepts across all science
  curricula
• Received an NSF CCLI grant (DUE 9952806)
• June 2000 - August 2003, 11 co-Pis to create
  educational materials for an undergraduate
  Computational Science minor
• Each course will be a collection of case studies
• Parallel and High-Performance computing will be
  integrated, in each course.

17
Courses to be developed

• Computational Science I and II
• Computational and Applied Mathematics
• Computational Biology
• Computational Chemistry
• Computational Environmental Science
• Computational Physics
• Computational Psychology/Neuroscience
• Scientific Visualization
• Parallel Computing (revised)
• High Performance Computing (revised)