Homework 1: Exploring Implicit and Explicit Parallelism with OpenMP

This project is a homework delivery regarding the first homework of parallel computing.

Table of Contents

1. Task Statement
   • Built With
2. Getting Started
   • Prerequisites
   • Installation
3. CSV Usage

Task Statement

In this assignment, you will explore both implicit and explicit parallelization techniques by imple- menting a matrix transpose operation. You will benchmark and analyze the performance of both approaches, comparing their efficiency and scalability. Consider a matrix M of size n × n, where n is a power of two. Implement the serial code, implict code and parallel code using OpenMP.

(back to top)

Built With

• 

(back to top)

Getting Started

⚠️ Be careful, this project has been written in C++ and compiled using gcc version 9.1.0, use the same compiler as I don't assure the same output results using another version. When using this project on the cluster make sure to have it in a folder, because the pbs file will create a build and logs directories, but upon starting it will also delete them.

Prerequisites

Ensure that you have installed gcc version 9.1.0

g++ --version 

Installation

1. Clone the repo

   git clone https://github.com/dimi56497/parallel-computing-H1.git

2. Copy the src directory on your HPC working dir

   scp -r src [email protected]:workingdir/

3. Copy the pbs file on your HPC working dir

   scp matJob.pbs [email protected]:workingdir/

4. Enter the HPC

   ssh [email protected]

5. Insert your working directory in matJob.pbs

   cd /home/username/dir; # Select your working directory

6. Run the pbs using the pbs scheduler

    qsub matJob.pbs

(back to top)

CSV Usage

The csv files contains:

• MatSize: size of matrix
• Time: execution time taken
• ThreadNumber: number of threads used (only present when using OpenMP)
• Valid: validity of data

(back to top)

Contact

Dimitri Corraini - [email protected]

(back to top)