Introduction to Supercomputer Applications
Grading 2nd Quarter Fall 2003

MPI Resources and Sample Programs

Introduction to MPI programming
Introductory Slides on MPI, by William Gropp
A Short Overview of MPI Concepts, by William Gropp
MPI Exercises, Pt.1

C Help Files - Sample Programs

Use of the Cray SV1 - resources, and Benchmarking programs for the Cray

2nd Quarter MPI/Parallel Programming and High Performance Computing Assignments

Proj. 5, MPI Program Lab03, IntervalSlicing.c - Calculate Pi, DUE Fri. Nov. 14, 2003
--parallelize the program "gregory.c"
Use MPI_Reduce() to sum data that is local to each process
Proj. 6, Lab04 Divide and Conquer with "Scatter" - collapsing integers
DUE Wed. Nov. 26, 2003
Proj. 7, Lab05 Parallel Mandelbrot - uses MPE graphics
DUE Fri. Dec. 12, 2003
Using Graphics in MPI: the MPE Graphics Library for MPI
Proj. 8a,b,c: Choose 3 programs from the following:
DUE Mon. Dec. 21, 2003
- Problems from "Parallel Programming" by B. Wilkerson - see handouts for more info
- Parallel Computer Vision - uses PGM image files
- Ping Pong Program
Proj. 9 Benchmark a program(s) on the Cray, Transformers, Elements
DUE Mon. Dec. 21, 2003
- Sample matrix program to run: matStarter.c
- Write and test a program to multiply two matrices
- Time some of your parallel programs with the Cray. See if there is
  optimization
- Benchmarking programs for the Cray SV1
- Computational Modeling programs
Final Projects, Proj. 10, 11 Choose 2 of the following labs - DUE Wed. Jan. 28, 2003
These are more advanced computationally intensive programming applications,
1. Holliman's Parallel Puzzlers:
  - Computing the Permanent of a Square Matrix
  - Sorting a List through a Heap Merge - a O(nlogk) algorithm
  - The Caravan Problem
  - Euclidean Travelling Salesman Problem
2. Introduction to Parallel Computing - Online Course,
  Course home page, UNC Charlotte Parallel Programming Group - Wilkenson and Allen
  For example:
  - Chap. 3 "Embarrassing" Parallel Computations
  - Divide and Conquer Parallel Algorithms
  - Pipelined Computations
  - "Assignment 3" - (2-dimensional) temperature distribution
  - Parallel Sorting algorithms, and Part 2 -Quicksort
  - Numerical Algorithms, Matrix Multiplication
  - MPI Lecture
3. Concurrent Wave Equation - Cornell Theory Center
4. Sharks and Fishes - Parallel Cellular Automata (from Univ. of Nevada, CS 732)
5. Class Projects from CS 732, Spr 2002, Univ. Nevada, Reno (fredh@cs.unr.edu), CS 732-Theory of Parallel and Distributed Processing
  - Proj 1 - Find max number in a file
  - Proj 2 - X-Windows mandelbrot.c
  - Proj 3 - Parallel Bucket sort
  - Proj 4 - N-Body astronomy problem
  - Proj 5 - Sharks and Fishes, cellular automata
  - Proj 6 - Shared Matrix Multiplication
  - Proj 7 - Shared Matrix Multiplication, pthread version
  - Proj 8 - Distributed Matrix
  - Proj 9 - Distributed Matrix Multiplication using a cluster
  - Notes from "Parallel Programming" by Wilkenson and Allen
6. Projects from CS 635 Foundations of Parallel Computing, GMU Fall 2002
  - Proj 1 - Parallelized "Game of Life"
  - Proj 2 - "Shortest Path
7. Designing and Building Parallel Programs, by Ian Foster
  - Parallel Algorithm Examples, from Chap. 1
    Exercises 6-10, Chap 1, require you to design a parallel algorithm
  - Case Study: Atmosphere Model, Chap. 2
  - Case Study: Floorplan Optimization, Chap 2
  - Case Study: Computational Chemistry, Chap. 2
  - Exercises for Chap. 2
  - Case Study: Shortest Path Algorithms, Chap. 3
  - Matrix Multiplication, Chap. 4
  - Exercises from Chap 4
  - Case Study: Earth System Model, MPI, from Chap. 8
  - Parallel Merge Sort, Chap. 11
  - Hypercube Algorithm exercises, Chap. 11
  - Designing and Building Parallel Programs, by Ian Foster, main page
8. Learn to use "Ray tracing" techniques for graphics realism
  --Write a POV-Ray program
  - NOTE THIS CHANGE TO THE POV-RAY command line:
    "x-povray ..." is changed to: "povray ..."
  - See Getting Started with POV-Ray
  - Animation with POV-Ray: write an animated sequence using POV-Ray
  - Try to utilize parallel programming and MPI to run an animated POV-Ray sequence of images
  - Use our Cluster: Working with the Linux Cluster POV-Ray Animation
  - Working with Remote Systems and More Notes on the Cluster
9. Using the TJ "Cluster": "Forking" processes on our Cluster
  Help for using the Cluster (the bank of computers in the Supercomputer room):
  - See Gentle Introduction on How to use Cluster Effectively by Evan Danaher
  - Ray Tracing and the Linux Cluster - see "Using the Linux Cluster"
  - About the Linux Cluster
  - Stats for the machines on the Cluster
  - To log on: "ssh cluster" - this logs you on to einstein
  - "mosmon" displays a graphic on the 16 processors
    This is a Mosix cluster, names of the 16 cluster computers
  - Try running a computationally intensive program on the Cray, the Cluster, and your student computer
  Fork program samples:
  - fork1.c
  - fork2.c
  - fork3.c
  - forktest.c, a variation of the program on Working with Linux Cluster and POV-Ray
10. Move on to MORE Parallel Programming Applications - Choose 1 or hopefully 2 from this group:
  - Use the Cray SV1: Test out our Cray supercomputer,
    - Run program(s) on the Cray
    - Compare the performance (using timing - MPI_Wtime()) with using MPI on our lab computers
    - Example Progams for comparison
      - calculatePi.c
      - calPiMPI.c
      - These two programs use the C function: gettimeofday(&tv, &tz);
    - Running programs on the Cray SV1
    - Information on Vector Architectures: our Cray is a Vector computer
    - EXTRA:
      - Use Vectorization processes and Optimization techniques
        See handouts:
        
        Introduction to Vectorization and Optimization, North Carolina Supercomputing Center
        
        Sample vectorization exercises: (c version), (fortran)
        
        Vectorization Sample Code (PDF) for the Cray, - PS version + "ex.c" the sample program
        "Performance Tuning for Vector Architectures" (PS - view with gv - ghostview)
        Optimization for Vector Computers(PDF - view with acroread), from Jim Giuliani's workshop
        Optimizing Code on our Cray
      - Optimize your MPI code
      - Performance Measuring Tools for the Cray: Guide to Parallel Vector Applications:
        
        FIRST: In order to view these tools -
        
        Sign on to the Cray
        In this x term, type:
        "setenv DISPLAY yourcomputer.tjhsst.edu:0.0"
        (Note, on a "bash" system, this is
        "export DISPLAY=yourcomputer.tjhsst.edu:0.0")
        Open another x term (logged on to your computer, not the Cray)
        In this second x term, type:
        "xhost + cray.tjhsst.edu"
        
        ATExpert
        Using the X Window Interface
        flowview
        Hardware Performance Monitor (hpm)
        jumpview
        perfview
        procview
        Hardware Performance Monitor Counters
  - Extra: Identify interesting Internet2 and Supercomputer application areas
    - Pick at least one application idea in each area (Internet2 and Supercomputing)
    - Describe why you think this is a good application area,
      what possibilities there are for creative and useful functions

Introduction to Supercomputer Applications Grading 2nd Quarter Fall 2003

Introduction to Supercomputer Applications
Grading 2nd Quarter Fall 2003