Colton Myers Kyle Rich Landon Gilbert-Bland

Red-Blue Computation

This project explores the red-blue computation that we discussed in class. It consists of sequential, OMP, and Pthreads versions of the computation in C++, a number of matrix files, and a Java version of the program that presents a visualization of the computation

All tests were performed on water.eng.utah.edu. Once logged into that server, the files can be compiled with the following commands:

make seq
make omp
make pthread

The following commands will run the resulting binaries:

./seq <size> <size> <filename> <converge_val> <converge_tiles>
./omp <size> <size> <filename> <converge_val> <converge_tiles> [num_threads]
./pthread <size> <size> <filename> <converge_val> <converge_tiles> [num_threads]

Note that num_threads is optional, and must divide the value evenly. Additionally, there is a threshold for convergance values for each grid above which the grid will not converge, or at least not in a reasonable timeframe.

An example run, using the 512x512 matrix provided, with a 16% convergence value and 8x8 tiles, would be executed with the following commands:

./seq 512 512 grid512x512.txt .16 8
./omp 512 512 grid512x512.txt .16 8
./pthread 512 512 grid512x512.txt .16 8

Another run, using the 4096x4096 matrix provided, with a 12.7% convergence value and 8x8 tiles, would be executed with the following commands:

./seq 4096 4096 grid4096x4096.txt .127 8
./omp 4096 4096 grid4096x4096.txt .127 8
./pthread 4096 4096 grid4096x4096.txt .127 8

Timing and iteration information will be automatically printed.