Fall 2006 CS 432/632/732 Parallel Computing
Homework-2
10/10/2006 1-2
Individual Work Only. 200 points. Due October 24, 2006.
Design and implement a distributed memory version of the image processing program
implemented in Homework-0 using MPI for message passing.
1. Use two-dimensional data distribution, test the program for different number of processes
shown in table 2, note down the time taken, and complete table below.
Table 2. Time taken (in seconds) with two-dimensional data distribution
Image1 Image2 Image3
No. of
threads Mean Median HP
Filter Mean Median HP
Filter Mean Median HP
Filter
Sequential
1 X 1
2 X 2
3 X 3
3 X 4
4 X 3
4 X 4
4 X 6
4 X 8
6 X 4
6 X 6
6 X 8
8 X 4
8 X 6
8 X 8
2. Using the execution time for image3, develop speedup and efficiency plots for varying
number of processes using two-dimensional data distribution.
3. Compare the execution time between the three different versions of the parallel programs
implemented (Pthreads, OpenMP, and MPI) and explain any difference in execution time
between the two versions.
General Guidelines:
Implement and test these programs on the CIS cluster – Olympus. Follow the instructions for
using Olympus at: http://www.cis.uab.edu/cs632/fall2006/OlympusInstructions.html.
Follow the examples in the MPI Tutorial and execute these examples before working on the
homework.
To minimize the number of job submissions, you could use a single script to submit different
images and different operations. For example you can use the following SGE script for one run:
