CIS 367: INTRODUCTION TO PARALLEL COMPUTING
Spring 1998
Programming Assignment 4

Due Date: Start of class, Tuesday, April 21, 1998

Objectives

The objective of this assignment is to code different data distributions of vectors and matrices in preparation for using these routines in the next assignment, and to increase your understanding of each distribution.

Procedure

1. Read through chapter 5 and 10.1 through 10.4 of the textbook.

2. Write an MPI program that takes two integer arguments on the command line, onedsize and twodsize, to be used to determine the size of the arrays to be allocated and used during this execution. Your program should declare a single dimension array called vector. Only process 0 should allocate space for this array, by using the command line argument onedsize and malloc. A two-dimensional square array called grid should also be declared for all processes. Again, only process 0 should create space the size of twodsize elements. Also, declare arrays local1d and local2d. In each process (including process 0), these arrays should be sized to hold (onedsize)/(number of processes) elements and (twodsize)/(number of processes) by (twodsize)/(number of processes) table of elements, respectively.

3. Write one routine that takes a single dimension array, its size, and a function of one variable as input arguments, and initializes the array using the function passed in as a function of the array index.

4. Write a routine that takes a two dimension array, its size, and a function of two variables as input parameters, and initializes the array using the function passed in as a function of the array indices.

5. For process 0 only, call the first routine to initialize vector such that vector[i] = 2 * i + 3; call the second routine to initialize grid such that grid[i,j] = w * i + j;

6. For all processes, call the initialization routines to set all elements of their local arrays to 0.

7. Write two print routines, onedprint and twodprint, each taking an array of the appropriate dimension, its size, and a rank. The print routines should print out the rank of the processor passed in and then a nicely formatted version of the array of the particular dimension. A vector should be labeled nicely and printed across the page, then on to the next line. A two dimensioned array should be printed as a table if possible. For a vector of length greater than 30, it should print out only the first ten and last ten elements, labeled as such. For a matrix of more than 20 by 20, it should print only the first 2 and last 2 rows, clearly marked as such.

8. Write a routine called distribute1d which takes as arguments: a single dimension array and all the necessary arguments to perform data distribution in a general manner including an argument that indicates the kind of distribution desired - either block, cyclic or block-cyclic. distribute1d should perform the correct data distribution for the number of processors and size of the array given as the argument. That is, it should be possible to perform a block, cyclic, or block-cyclic distribution of any size 1-dimensional array for any number of processors, by giving the correct arguments to distribute1d.

9. Write a similar routine called distribute2d that is capable of data distribution of a 2-d array for (block,*), (*,block), (cyclic,*), (*,cyclic), (block, block), or (cyclic,cyclic). Recall that (block,*) is block-row and (*,block) is block-column.

10. In your main program, write calls to these routines to accomplish each of the listed distributions above, in the order given above for grading purposes.

11. Between each call to the distribution routines, each process should call the appropriate print routine to print out their rank and their local array values. Then, the appropriate initialization routine should be called by all processors to initialize their local array variables to 0 in all positions.

12. Compile and run the program for 1, 4, and 8 processes, each with command line arguments:

    32 32
    64 64
    128 128
    256 256

    So, there should be a total of 12 runs. You do not have to do any timings!

Experimental Report

Your experimental report should consist of the following sections in this order. It is strongly recommended that you type your report using a word processor rather than handing in a hand-written report.

1. Cover Page:

   Title, author, course number and semester, date.

2. Project Summary:

   In one paragraph, summarize the activities of the lab.

3. General Comments:

   This section consists of a discussion of what you learned from the lab. Discuss your view of the issues involved in implementing each of the distributions.

4. Appendix:

   1. Your code for the complete program.
   2. Your output from each run of the program, clearly marked. A script showing your compilation and mpirun command lines and output is preferred.

About this document ...

This document was generated using the LaTeX2HTML translator Version 95.1 (Fri Jan 20 1995) Copyright © 1993, 1994, Nikos Drakos, Computer Based Learning Unit, University of Leeds.

The command line arguments were:
latex2html -split 0 lab4.tex.

The translation was initiated by Lori Pollock on Tue Apr 14 14:06:27 EDT 1998