Exploiting hardware parallelization to optimize Matrix Multiplication problem

**Tharindu Rusira, Chalitha Perera**
*Department of Computer Science and Engineering*
*University of Moratuwa*
*Sri Lanka*

---

## Compiling the code

This project uses Cmake utility to compile the code and build necessary libraries. Download the repository, navigate to 'build' directory and run 'make'.

If everything goes well, an executable file with the name "lab4" will be generated.

Executing Compiled Program

In order to run compiled code you need to give arguments matrix multiplication version and the dimension of the square matrix.

Running Serial Version with dimension 1000

./lab4 serial 1000

Running Parallel Version with dimension 1000

./lab4 parallel 1000

Running Optimized Version with dimension 1000

./lab4 optmized 1000

Determing Number of iterations to get runtimes within an accuracy of ±5% and 95% confidence level

use lab4_stats.py script

example

python lab4_stats.py --run_command="./lab4 parallel 1000" with execute matrix multiplication with

default of 5 iterations and give average, standard deviation and required number of iterations

then again run with n iterations to get the average

python lab4_stats.py --run_command="./lab4 parallel 1000" --iterations=n