CS 320/CSE 302/ECE 392
Introduction to Parallel Programming for Scientists and Engineers
 
Machine Problem 1

Assigned: Tuesday, February 8, 2000
Due: Friday, February 18, 2000

The objective of this MP is to transform traditional loops of a Fortran/C program into an equivalent parallel form using the OpenMP extensions.

The development machine is newton.cse.uiuc.edu. The OpenMP implementation from KAI may be found in /home/cse/cse302/cse302/class. Add the path /home/cse/cse302/cse302/class/bin to your PATH environment variable. For convenience add it to your .login (something like: set path=($path /home/cse/cse302/cse302/class/bin)). The compiler is called guidef77/guidec for Fortran 77 and C respectively. A debugger for OpenMP is also present: assuref77 (Fortran 77) and assurec (C). The manuals can be found in the directory /home/cse/cse302/cse302/class/docs. To invoke the compiler/debugger use: <compiler/debugger> <program_file> which will produce an a.out file. Further usage can be found in the manuals.

For this MP we shall use the program mp1.f (Fortran)/mp1.c (C) which can be found in /home/cse/cse302/cse302/class/mp1. Both represent the same program in two different languages. You can work on whichever version you prefer.

The program is an implementation of the LU decomposition of a matrix (for non-singular matrices) alongwith a permutation matrix. This is then used to find the solution to the standard linear equation A·x = D . You have to convert the loops to a parallel form. The credit you receive will be in accordance with the number of loops you parallelize.

Instructions for handing the MP will be posted on the web page shortly.


File translated from TEX by TTH, version 2.61.
On 8 Feb 2000, 12:34.